《人工智能小白书》（第 2 部分 · 第 13–16 章）

The Little Book of AI (part 2, chapters 13–16)

> By Li Xiaolai · his book manuscript published on GitHub (his own Chinese and English versions) · [GitHub repository](https://github.com/xiaolai/little-book-of-ai) · version of 2026-10-11. Images are given as links.

07. AI Doesn't Tolerate Laziness

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/dalle/07.png)

7.1. Lazy People Drive the World?

Let's kick things off by discussing a popular notion floating around. Some folks claim that the world is driven by lazy people. There's even a catchy saying:

> *If necessity is the mother of invention, then laziness is sometimes its father.* Yes, if we look back, we'll find that laziness is behind many inventions. Laziness drives innovations that boost productivity because people often think, "There must be a better way."

Clearly, this is just a flashy statement, yet it has quite a following, both at home and abroad. They often cite a quote from [John Atanasoff](https://en.wikipedia.org/wiki/John_Vincent_Atanasoff), the inventor of the electronic calculator, as evidence:

> "I was too lazy to calculate and so I invented the computer."

People who can't see the forest for the trees think they've stumbled upon a profound truth. They get all excited and start listing a bunch of what they believe are solid examples:

> - Too lazy to climb stairs, so they invented elevators.
> - Too lazy to walk, so they invented cars, trains, and airplanes.
> - Too lazy to cook, so they invented microwaves.
> - Too lazy to attend concerts, so they invented records, tapes, CDs, and MP3s.
> - And the most outrageous one: too lazy to kill one by one, so they invented the atomic bomb...
>

Surprisingly, even some famous people support this flashy claim and agree with this nonsensical reasoning.

Where's the flaw in this argument?

The truth is clear: the root of all inventions is definitely not because inventors are "lazy in every aspect." Take John Atanasoff, for example. He wasn't lazy in every way. He was just tired of doing tedious calculations. In other, more important areas, he was anything but lazy. For him, saving time on boring calculations meant he could do more meaningful work, like inventing new calculators.

As for the so-called "evidences" these overly excited people present, it's not worth refuting. It's unnecessary and exhausting because it's all over the place. The people too lazy to climb stairs aren't the ones inventing elevators. The inventors of cars, trains, and airplanes weren't those too lazy to walk. The inventors of microwaves weren't necessarily too lazy to cook. Who says the inventors of records, tapes, CDs, and MP3s were too lazy to attend concerts? It's baffling how they can establish such random cause-and-effect relationships and then take them for granted.

Even those described as "too lazy to kill one by one" don't stop killing just because they have atomic bombs. If they're serial killers, mass murderers, or those who start wars without reason, they're actually quite diligent and hardworking in the act of killing, aren't they? Where's the laziness?

Lazy people can't drive the world. If someone is truly lazy to the core, they wouldn't bother changing the world, right?

Plato once said in "The Republic," "Necessity is the mother of invention." But do we really need to find a partner for this "mother"? Claiming "laziness is the father of invention" and feeling smug about it seems like a puzzling performance art to me. However, if we tweak Plato's saying to "Efficiency is the mother of invention," no one would "seek attention" because "efficiency" has never been a lazy person's preference.

7.2. The Meaning of Tools

Why start by wrestling with this question? Because it touches on our fundamental perspective on using tools, a crucial concept in the methodology of tool usage. What are tools really for?

> Every tool (or invention) is born to **boost efficiency**, but using tools isn't just about making things easier.

At first glance, this might seem a bit tricky because sometimes "making things easier" and "boosting efficiency" appear synonymous. However, let me break it down from another angle. The primary purpose of using tools isn't just to make things easier; it's to boost efficiency.

Every task, every job, and every meaningful endeavor in the world often consists of 99% **tedious and monotonous yet essential** components. The parts people consider advanced might only make up 1%.

The renowned writer Bi Shumin mentioned a brilliant analogy in her novel "The Female Psychologist":

> A cup of coffee is 99% water, with the actual coffee making up just 1%.

If you visualize that 1%, it's just one square in a 10x10 grid, as shown below:

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/1-percent.png)

> In the past, drawing a diagram like this was quite a hassle. Now, it's simple. I just tell ChatGPT:
>
> Draw a svg, 10x10 grid, with top-left corner filled with coffee.
>
> https://chatgpt.com/share/6706000a-4b1c-8009-9995-6dbbaefb306a
>
> And I get the result almost instantly. As for what `svg` is, you can learn about it by asking the [Learning Anything](https://chatgpt.com/g/g-dPqJT0PXS-learning-anything) bot...

"Having ChatGPT draw this diagram for me" is a great example. I did make things easier by using it, and this ease did boost my efficiency. But the key point is that "drawing this diagram" is part of the 99% **tedious and monotonous yet essential** tasks in my current project — writing a book. It's not part of the truly important 1%. This diagram is nice to have, but not having it wouldn't be a dealbreaker.

So, here's the crux:

> All tools, whether "smart" or not, often can't do much for the most important 1%. Their real value lies in efficiently tackling the 99% **tedious and monotonous yet essential** tasks.

If tools can help us efficiently handle the high-percentage **tedious and monotonous yet essential** tasks, then we have more time and energy to focus on the important 1%. We might even increase its proportion to 2% or 3%, right?

Ultimately, the role of all tools is similar. The best tools are often most effective in the 99% tedious and monotonous yet essential parts, while the truly important 1% inevitably relies on soulful, wise humans. This is an unavoidable reality and a long-standing fact.

So, contrary to what many people think, or even the opposite:

> Having better tools doesn't mean "less work to do," but rather "*more* work to do."

"Making things easier" is synonymous with "boosting efficiency." The result of boosting efficiency is, of course, "completing tasks faster." But what happens after "completing tasks faster"? Clearly, it's to "complete more tasks," not fewer. This is the full purpose of all inventions. Seeing only "making things easier" without seeing "more tasks" is a one-sided view.

Why do most people's thoughts contradict "facts"? Strange, right? Not really. People perceive "facts" differently. This cognitive difference in "tool methodology" comes from the tool users themselves.

As described in Chapter 5, most people are lifelong "program executors" rather than "program designers" or "program optimizers." Thus, whether it's work or tasks, what falls into their hands is often not the "important 1%," but more likely the "tedious and monotonous yet essential 99%." So, in their world, "facts" are just that, "tools are for making things easier," and another "fact" is obvious and real to them: "the better the tool, the less work there is."

> Think about it: the widespread discussion of "AI causing massive unemployment" and the resulting anxiety are likely based on the same "fact."

In another world, tools aren't just for "making things easier." The result of having better tools is that the work you can do and need to do only increases, not decreases. Not surprising, right? Because good tools help us tackle more tedious and monotonous yet essential tasks. Thus, the following phenomena become common:

> - What was originally important becomes even more important.
> - Important parts that were previously neglected can now be addressed.
> - We always discover more important parts...
>

The end result is that once you have better tools, you'll find more meaningful things to do. With better tools, people don't become more idle; they often become busier.

Non-learners, non-producers, and non-creators can't grasp the essence of truly doing things. They not only don't understand that 1%, but they aren't even aware of its existence. They uniformly and consistently believe that 99% is everything, and they spend their entire lives on that 99%. They share a common trait: when they're at a disadvantage, they never realize it, and they can't imagine why they're at a disadvantage. As a result, they spend their lives continuously at a disadvantage, suffering greater losses... Ultimately, the only "compound interest" they enjoy is the scenario of "being at a disadvantage."

If you're not a learner, producer, or creator, the same tool in your hands can have completely opposite uses and effects. Isn't that fascinating?

The internet celebrity mentioned in Chapter 2, who made three mistakes in one sentence, or his copywriting team, also suffered from this — they didn't do the most important thing. Let me ask, isn't fact-checking supposed to be part of that "important 1%"? With a good tool in hand, they only focused on "making things easier," mistakenly thinking that 99% was everything...

7.3. The Best Way to Be Lazy

I learned early on that "laziness" is something you absolutely can't cheat. It's like it has a mind of its own, holding grudges. If you try to cheat it, it won't say anything right away, but it'll sneak back and take more from you. No one has ever beaten it. So, I have a deep respect for laziness and never dare to offend it. The best way to be lazy is simply not to be lazy — never be lazy.

If you're a parent, you should show your kids the "composition of work or tasks" early on — the "key parts" and the "tedious and monotonous yet essential parts." The balance is heavily skewed, with the former being a small portion and the latter a large one. Observe and demonstrate this repeatedly in life. All work, all products, and all significant achievements are composed this way.

At the same time, instill the correct "tool methodology" in them: when you have a better tool, it can help you quickly and efficiently handle those tedious and monotonous yet essential tasks, and that's worth celebrating. However, the most important part, though smaller in proportion, requires your own time and brainpower, and you absolutely can't be lazy about it.

This world is full of lazy people. They make up a larger portion and will significantly increase in the future. So, in the foreseeable future, many funny things might happen or have already happened. For example, many people now rely on AI to write work emails. They do this to save effort, as AI can quickly expand any short phrase or keyword into any length — 150 words, 500 words, 2000 words, 3000 words — it doesn't really matter.

Meanwhile, on the other end, the person receiving the email no longer reads the full text themselves but relies entirely on AI to generate a summary. They then only read that summary, compressing the other person's lengthy text back into a sentence or even a keyword.

The final scene is quite amusing: it seems like there's a lot of communication between humans, but in reality, it's just two bots burning through tokens, computing power, and electricity, achieving nothing. Human communication, already sparse and inefficient, is now almost nonexistent, both in content and efficiency.

Lazy methods should never be emulated. For instance, the phenomenon we mentioned in last chapter about lazy people using tools: lazy people never spend time and effort opening the settings dialog to carefully study each option. Almost 99.99% of people won't do this. Therefore, the way lazy people use tools should never be emulated, or you'll gradually evolve from lazy to clumsy. This isn't a joke or harsh; it's a simple and clear fact.

7.4. Reading and Thinking

Since ChatGPT came along, the browser plugin market has been flooded with all sorts of "Summarizers," like mushrooms after a rainstorm. They're all paid, and surprisingly, they sell well.

I haven't installed a single one.

It's not because I'm stingy. It's just that "that's not how I use tools."

When it comes to "reading," people have a lot of misconceptions. The root is pretty straightforward. The fact is, most people don't read books; among those who do, many don't read anything substantial; and among those who read substantial books, a significant portion doesn't finish them, let alone truly understand them. Ultimately, for the vast majority, "reading" and "understanding" are two separate things.

So, it's quite amusing and even absurd: if you seriously read substantial books, finish them, and understand them, you've already outpaced the majority—saying it's 99% isn't an exaggeration. Throughout this book, I'll repeatedly emphasize this fact: outpacing 99% isn't hard.

7.4.1. The Professional Way to Read

Over the years, the process of "reading" has undergone significant changes, each driven by technological advancements that have turned things upside down.

Not too long ago, serious readers used "catalog cards" to take notes — not notebooks (that's amateur), but cards (that's professional). These cards would pile up and eventually require dedicated drawers for organized storage. Today's kids might not have seen these cards at all, or maybe only in some old libraries, because "technological progress" wiped them out years ago.

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/catalog-cards.png)

The "Catalog Card System" is actually a modern invention. Before [1780](https://web.archive.org/web/20220302063835/https://www.onb.ac.at/en/about-us/650th-anniversary/timeline/1780-the-oldest-card-catalogue), the world didn't have this system. After that, libraries everywhere relied on it.

I was lucky as a kid because my mom worked in a library. So, I learned to take reading notes the "professional" way early on — after all, those cards were everywhere at home, ready to be used. Notes on cards could be categorized and searched, making it easy to find what you needed later and avoiding the "I've seen this before, but where?" dilemma.

I used cards for note-taking until the late '90s. But then I stopped. Why? Because taking notes on a computer became more convenient, especially for searching. No matter how many cards you had, you could get search results instantly. Plus, typing is three to five times more efficient than handwriting!

Around 2005, things got even better as the number of e-books started skyrocketing. By 2010, a significant portion of the books I bought on Amazon were e-books. By 2015, almost 100% were digital. From then on, I didn't even bother with note-taking — no need to type out notes. As long as I had a general impression in my mind, I could search the full text later and find what I needed instantly among my saved books.

To this day, having a full-text search software on your computer is indispensable for serious readers. On macOS, I use [Devonthink](https://www.devontechnologies.com/apps/devonthink), the professional version at $199. For someone who takes tools seriously, that's not expensive. After all, I was the kind of person who spent five or six hundred bucks on a screwdriver in the '90s. There are free alternatives, like [DocFetcher](https://docfetcher.sourceforge.io/en/index.html) (which also has a "professional" or commercial version). While it might lack some features, it's cross-platform, available on macOS, Windows, and Linux.

This is a classic example of how "better tools" gradually eliminate the time-consuming and tedious yet essential parts — like using catalog cards for reading notes. It certainly makes things easier, but what's the result? The result is definitely "more time and energy to read more books," not "no need to read anymore" or "spending all the extra time on leisure," right?

7.4.2. Reading Summaries is a Waste of Time

With ChatGPT now in the mix, it can whip up a "summary" of just about anything in no time. But here's the kicker: is that really useful? It's like back in the day when I read books and took notes on cards, and you, trying to save time, just skimmed through someone else's cards — does that really work?

"Reading" and "reading summaries" are worlds apart. Often, reading summaries isn't just unhelpful; it's a total time-waster. This is especially true for important knowledge and skill-based books. If you're trying to learn programming, what's the use of a programming book's summary? No matter the language, the summaries and tables of contents are pretty much the same. If you're learning English, what's the point of a grammar book's summary? Regardless of the publisher, the summaries and contents are basically identical.

As an author, I often think, "If you can understand what I've written just by glancing at the outline or summary, why did I even bother writing the book?" Conversely, if you can truly grasp everything just by skimming the outline or summary, it means I've been "padding" the entire book beyond the outline!

Even for entertainment, like novels or movies, you can't just read the summary! Take Jiang Wen's "The Sun Also Rises" — if you boil it down to a summary, it's just one sentence: "A 'most lovable person' messes around and ends up accidentally shooting his own son." What's the point? "Breaking Bad" is a masterpiece, but summarizing five seasons into one sentence—"How bad can a good guy get?" — doesn't do it justice. Is that interesting?

Many people might be misunderstanding. "Reading" isn't just an "input" process; it can actually involve two or even three stages:

```mermaid
flowchart LR
input(Input) --> processing[[Processing]] --> output(Output)
```

If "reading" were only about "input," then having ChatGPT summarize for you would indeed "boost efficiency." But if the second stage is more crucial, where you need to process, i.e., think... then what's the use of a summary? Thinking is clearly your job, not AI's. The so-called "reading without understanding" often isn't due to a lack of brainpower but because there's only the first stage, with no second stage, leaving your brain with nothing to do... And that's not all; there might even be a third stage! Without the second stage, and with ChatGPT handling the first, what do you have to output? How can you output anything?

Among the stages of reading, the most important is "processing," i.e., the thinking we do while reading. Saying this stage is "important" is an understatement; it's downright "indispensable." Without this stage, can you even call it "reading"?

7.4.3. Thinking While Reading

When you're diving into a book, there's a lot of brainwork involved.

If you're into fiction — novels, movies, TV shows — beyond just getting hooked on the plot, you might notice how the author shapes characters, the actors' performances, storytelling techniques, and even the deeper meanings behind the story.

For nonfiction, you're evaluating the author's arguments, checking their logic, methods, and the authenticity of their evidence or data. Plus, you need to dig into the assumptions hidden between the lines or behind the text. And then, the content might spark countless associations in your mind...

When it comes to skill-based books, you're constantly figuring out things you didn't understand before, memorizing essentials, organizing your thoughts, and practicing until you've truly mastered the material.

All this thinking happening in your brain is something no tool can do for you. Frankly, it shouldn't. Tools exist to tackle the tedious and monotonous yet essential tasks, not to replace your brain or rob you of your thinking.

Growing up, teachers often said, "Read with questions in mind." But they rarely told us what questions to ask. That's where ChatGPT can step in—if you read its responses carefully, you'll find it incredibly useful:

> For each question below, you can "follow up" with: "Translate the above response into Simplified Chinese."
>
> * **Reading Nonfiction Books**
>
> > Provide a comprehensive and structured list of ChatGPT prompts that can guide a reader through the process of reading a nonfiction book.
> >
> > https://chatgpt.com/share/67062737-5528-8009-9936-1fdc6d431269
>
> - **Reading Skill-Based Books**
>
> > Provide a comprehensive and structured list of ChatGPT prompts that can guide a reader through the process of reading a book for skills or techniques.
> >
> > https://chatgpt.com/share/670627ab-6dd0-8009-aac2-53687440ea2f
>
> - **Reading Fiction**
>
> > Provide a comprehensive and structured list of ChatGPT prompts that can guide a reader through the process of reading a fiction.
> >
> > https://chatgpt.com/share/67062822-5adc-8009-9e55-f72981061d76
>
> - **Reading Poetry**
>
> > Provide a comprehensive and structured list of ChatGPT prompts that can guide a reader through the process of reading a poem.
> >
> > https://chatgpt.com/share/6706289b-1bb0-8009-958a-7b0f84abc5eb
>
> - **Watching Movies and TV Shows**
>
> > Provide a comprehensive and structured list of ChatGPT prompts that can guide a reader through the process of watching a movie or TV series.
> >
>
> https://chatgpt.com/share/670628f5-77b8-8009-9364-85f3a29e6ae3

The benefits of reading can be measured. In "The Truth About Thinking," I laid out a simple and clear standard:

> At any time, all effective input will inevitably cause changes or optimizations in the three tables in your mind (i.e., your operating system):
>
> > * Concept Table
> > * Relationship Table
> > * Program Table
>
> —— If these tables don't change after input and processing, it means the input was ineffective.

The thinking process during reading is one of life's rare and immensely enjoyable experiences. I don't rely on "summary tools" or "abstract tools" for a simple reason: I won't let them take away my reading joy, or more precisely, the *joy of thinking while reading*. Even for video summaries, I don't use them because my stance is: if I'm going to watch, I'll watch seriously; if not, I won't watch at all. For example, I revisit amazing shows like "Breaking Bad," movies like "The Yellow Sea," or classic novels like "Fortress Besieged" every year. I not only take them seriously but also enjoy them repeatedly, and each time is a delight.

The thinking process during reading is like a long-distance run; it's truly the brain's "best aerobic exercise." You'll eventually realize that another important aspect of reading is not just gaining knowledge but also strengthening your brain. Although this process isn't always obvious, it's similar to physical fitness. A healthy body enables you to do more and boosts your confidence and charm. Similarly, enhanced intellect can have the same effect, possibly to an even greater degree.

7.4.4. The Book Genie

In the first chapter, we used an analogy:

> If Google is like a "giant library," then ChatGPT is like a "genie" that suddenly appeared in an even bigger and higher-quality library. It's constantly expanding and organizing the library while having a clearer understanding of everything inside. And since it's also an AMA bot, it's always ready to have a normal conversation with anyone at any time.

You can literally use ChatGPT as a library. It's truly a book genie.

The reason those "summary generators" out there are useless is this: as long as a book is "legit," a good book, and has been reviewed by many, ChatGPT essentially has that book "built-in." You don't need to "upload files"; just start a serious conversation about the book with it.

Think about it. How often do we actually get the chance to read a book that was published just yesterday? A huge chunk of books, especially serious and classic ones, were published years ago. So, there's no worry about ChatGPT being "outdated" in this regard.

During the 2024 National Day holiday, a friend recommended [The Creative Act: A Way of Being, by Rick Rubin](https://www.amazon.com/Creative-Act-Way-Being/dp/0593652886), published on January 17, 2023, and a New York Times bestseller. Of course, I didn't bother with any summaries. I spent an entire afternoon slowly reading it. However, when I messaged ChatGPT in the same way:

> Provide a comprehensive and structured list of ChatGPT prompts that can guide a reader through the book "The Creative Act: A Way of Being."
>
> https://chatgpt.com/share/67062fc9-bae4-8009-a025-c0e525e1029f

Clearly, ChatGPT knows the book well. Since I had already read it, thought about it, and made connections, I could answer most of the questions myself. The surprise was that the structured list included several questions I hadn't considered — pretty cool!

I had completely abandoned the habit of taking reading notes. For me, writing books is my "study notes" or "reading notes" — except they're things I can sell, so I invest more time and effort.

However, I've started taking notes again—because there are so many questions worth asking AI, and often parts worth saving. Besides using ChatGPT, I frequently use Google's [NotebookLM](https://notebooklm.google.com/), launched in January 2023. It offers a lot of functionality. By September 2024, it could generate a 15-minute podcast in conversation form about any book you throw at it.

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/NotebookLM-google.png)

Of course, in NotebookLM, I ask the same questions in the same way:

> Create a structured list of prompts that can guide me through this book.
>

Both NotebookLM and ChatGPT accept user-uploaded text. Currently, neither directly accepts `epub` format e-books. So, for any book, I usually upload them in `txt` format. To convert `epub` or other formats to `txt`, you can simply ask ChatGPT...

> How to convert epub into txt?
>
> https://chatgpt.com/share/67063a4f-afec-8009-a732-e1a2ccb8f308

For basic computer-related knowledge like this, ChatGPT is unlikely to have any "hallucinations" and can be used right away.

For over twenty years, most of the books I've read have been "non-fiction." I rarely read novels, opting instead for lots of movies and TV shows. When I encounter a great piece, I become quite meticulous. I don't want to miss any detail, and I always use search engines to find more interesting stuff. With AI, I'm even happier because now I can directly ask questions like:

> List as many as possible *less known interesting things* in Breaking Bad.
>
> https://chatgpt.com/share/6706383c-4b48-8009-ba35-6358d9d3a775
>
> (And don't forget to run it by @Fact Checker...)

With AI, not only has my viewing experience become more diverse, but I've also picked up novel reading again — because being able to directly interact with a novel is a magical experience. If you want to know how to play with novels using AI, you don't need me to tell you. Just ask ChatGPT — and remember to ask several times:

> List as many as possible creative ways talking to a novel or fiction using AI.
>
> https://chatgpt.com/share/67063919-2c94-8009-a282-25ca7664d3b7

These are all great examples, each proving that "with better tools, the things you can and want to do only increase..."

7.5. Never Give Up Thinking

This chapter is a bit of a curveball. In a book about using AI, here's a whole chapter urging you not to use AI in one specific situation: when you need to think for yourself.

We've been saying it all along—"don't give up thinking," or "stick to independent thinking." We flagged this back in Chapter 2:

> You can ask ChatGPT, but at the same time, **maintain independent thinking**. Filter through ChatGPT's responses and add your own insights...

— But chances are, you might not have noticed or taken it seriously back then.

Earlier, we talked about how the world is full of lazy people, and their numbers are only going to grow. But a scarier trend has already emerged: people who are unwilling or simply refuse to use their brains. With AI's rise, their numbers and proportions are skyrocketing, and they're doing it with pride.

The human brain, from birth to death, is highly plastic. Technology has a massive impact on it. Word processors with auto-correct have made many people unable to write correctly with pen and paper. Navigation apps have left a huge chunk of people unable to easily identify directions, relying instead on left and right. Smartphones have shortened everyone's attention span to an alarming degree. Many people can't even sit through a movie anymore and can only handle short videos. Some can't even watch short series and go straight to ChatGPT for summaries.

That's how the brain works. The moment you stop thinking, it loses its ability to think, with no regrets or second chances.

Research shows that identical twins, though genetically identical at birth, can develop vastly different personalities and paths. Often, one is "decisive," while the other is "indecisive"; one is "extroverted," while the other is "introverted." A key reason behind this is "brain plasticity."

This applies not just between people but also between humans and robots. The human brain can "reshape" itself rapidly, unlike "robot brains."

The greatest value of AI is probably freeing people from "boring and monotonous yet essential" tasks, allowing them to focus their limited attention on more meaningful activities like learning, producing, and creating. Whatever you do, it requires "focused thinking."

At a pivotal moment in AI, an important paper titled *[Attention is All You Need](https://arxiv.org/abs/1706.03762)* was published. Here, "you" refers to AI algorithms. If we borrow this phrasing for humans, it should be *Attention is All You Have.* Here, "you" refers to us. Since attention is our only asset, our most tangible one, it makes sense to invest it in thinking.

But what's happening? More and more people don't care about their only asset, scattering it carelessly, with no regrets. As for the most valuable activity — thinking — they're handing it over to AI, cheering all the while... It's both sad and unfortunate.

In the end, it's a choice. And when it comes down to it, you have to own the choices you make.

08. Disrupting the Learning Path

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/dalle/08.png)

In September 2023, OpenAI released the third version of DALL·E, and it caused quite a stir. The name DALL·E is a mashup of two sources: the 2008 Pixar movie, *WALL·E,* and the Spanish artist *Salvador Dalí*. It's pronounced `/ˈdɑː.liː/` in American English or `/ˈdɔː.liː/` in British English.

> https://chatgpt.com/share/67066451-9d2c-8009-9a1f-55e79f5f1701

DALL·E is pretty fun to play with, but let me digress for a moment and talk about something else first.

8.1. A Problem That Used to Bother Me

I've always had a beef with those "board books" designed for kids aged 0-3. To me, they're pretty much a complete waste, a classic case of a fake need.

From a design perspective, the bright colors, simplified shapes, and cartoonish images don't really do much, even though we get why they're designed that way.

The idea that "bright colors attract kids' attention" is mostly a parental illusion. From a visual neuroscience standpoint, while bright colors can grab a kid's attention momentarily, that attention is fleeting and easily distracted.

Cartoonish designs are similar. They simplify visual information, helping kids quickly recognize objects or characters, but they don't do much for long-term cognitive development. Research shows that more realistic images or real-world objects actually help kids understand the world and build real conceptual connections. The brain prefers content that aligns with real experiences, and overly simplified cartoon images can sometimes distort a child's understanding of the real world.

Kids in this age group are more inclined towards "overt attention," meaning they react quickly to obvious, standout stimuli in their environment. They're drawn to dynamic stimuli like sounds or moving people and things. Compared to these, bright colors, simplified shapes, and cartoon images are pretty insignificant in capturing a child's attention.

Kids aged 0-3 can't grasp the concept of a "book" for quite a while. In their world, it's just another household item to throw, tear, or even chew on. It's definitely not what those parents, eager for early "educational enlightenment," imagine it to be.

Board books are really just there to cater to these parents — the main selling point being that they can't be easily destroyed by little hands. Of course, eventually, they'll get destroyed in all sorts of unexpected ways.

But are they really any good? Is the ink and cardboard used in printing safe? Are they free from all toxic substances? From what I've seen, most parents are diligent about washing their kids' hands but seem unconcerned about the cleanliness of those board books.

The numbers, letters, colors, and shapes printed on them perfectly cater to parents' desires for early "educational enlightenment." However, kids aged 0-3 actually need three other things:

> * Interaction
> * Language exchange
> * Life experiences

And these are precisely what board books and their content can't provide. Besides having poor content, these things are ridiculously expensive — priced to exploit parents' willingness to pay for "early education," not based on the actual value of the books.

Honestly, the most frustrating part isn't just that they're useless, overpriced, and possibly unsafe. It's knowing they're junk but feeling powerless to do anything about it. Letting a piece of junk take up a frustratingly important place in your life is as maddening as dealing with a lousy boss you have to endure at work.

On top of that, I have another gripe.

I'm ethnically Korean, grew up in China, and have lived in a bilingual environment with both Chinese and Korean as my native languages. I later learned English, so I'm basically a polyglot. I know the lifelong benefits of multilingual skills and the importance of a multilingual environment.

So, while I dislike board books, I'm not entirely against picture books. In fact, I wish those picture books were in English.

However, even English picture books have their issues — not everyone at home speaks English. The text in the picture books is often silly, but that's secondary. The main problem is that family members can't narrate them; even if they could, they can't do it in English. Making up stories on the fly is tough, and even if you prepare, it doesn't flow well. I can't draw or make up stories, let alone do it continuously. Neither can my family — it's a mess.

So, I just stay frustrated. Then I try to forget about it and move on.

8.2. Solutions with AI

Suddenly, combining DALL·E, ChatGPT, and TTS seemed like a solid solution! So, I asked ChatGPT:

> Help me draft a plan to create a picture book for toddlers around 2 years old using DALL·E, ChatGPT, and TTS. List the steps in detail.

ChatGPT eagerly generated *several* versions based on my requests. Here's a summary of one of the more satisfying responses:

> **Plan Steps**
>
> 1. Define the story theme and content
> 2. Use ChatGPT to write the story text
> 3. Generate illustrations with DALL·E
> 4. Combine text and images
> 5. Use TTS to create an audio version of the story
> 6. Integrate audio and images to make an e-picture book
> 7. Test and optimize
>
> https://chatgpt.com/share/670671e8-9b04-8009-812f-4d5149cb9dcd

After a quick read and understanding, I decided to start with a minimalist version:

> * DALL·E: Generate just one image;
> * ChatGPT: Create a short English story or description for the image;
> * TTS: Convert the story text into English audio.

And just like that, I got to work.

8.2.1. First Attempt

Remember our "prompt optimization template"?

> Please help me refine the following prompt according to best practices for prompt design, so I can get the best feedback from ChatGPT. Return the modified prompt in both English and Chinese. Also, please do not use Markdown code blocks in this response:
>
> [Enter the prompt you need to modify again]

I've saved it in the Prompt Library Manager plugin, ready to copy and paste whenever needed. Another option is to use a clipboard manager to keep this frequently used template handy.

I opened the Edge browser in split-screen mode, with two chats side by side. The right side was for optimizing prompts, and the left side was for chatting directly with ChatGPT.

[〔图：Edge Split Screen〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/edge-split-screen.png)

On the right side, I casually typed (but within the template):

> Help me generate an image suitable for a 2-year-old.

The optimized prompt on the right was copied and pasted to the left, with a few tweaks:

> Generate an image suitable for a 2-year-old. The image should be colorful and friendly. Use soft lines and bright colors to capture a toddler's attention. Ensure the image is full of positive energy, with a warm and cheerful atmosphere.

No need for `@DALL·E`; ChatGPT now generates images directly with such prompts. The resulting image was okay—not exactly what I wanted, but it worked.

On the right side, I casually typed again:

> Please write a short story in American English, around 50 words, based on this image.

Once more, I read the optimized prompt, copied it from the right side to the left, trimmed a bit, and sent it to ChatGPT:

> Write a short story in American English, about 50 words, based on this image. Use simple language suitable for toddlers, ensuring the story is lively, engaging, and full of positive energy to spark a child's imagination.

I quickly got a response:

> Once upon a time, a little bunny named Benny lived in a sunny meadow. Benny loved to hop and play with his best friend, Tilly the turtle. One day, they found a shiny red ball! They kicked, tossed, and giggled all day long. "Friends make everything fun!" said Benny, smiling at Tilly. And they both laughed.
>
> https://chatgpt.com/share/67068074-4494-8009-8c43-f1a0ccae4ea7

I read it and felt something was off... Where did Tilly the turtle come from? Plus, there was no shiny red ball in the image!

The takeaway from this attempt is that ChatGPT's "image reading" skills are still lacking. They'll definitely improve over time, but for now, it's not quite up to the task.

8.2.2. Second Attempt

Let's switch gears and try generating the story first, then create an image based on the story. This is a kind of "workflow optimization," right?

Original workflow:

> 1. DALL·E: Generate one image;
> 2. ChatGPT: Create a short English story or description for the image;
> 3. TTS: Convert the story text into English audio.

Revised workflow:

> 1. ChatGPT: Generate a short English story or description;
> 2. DALL·E: Create an image based on the story;
> 3. TTS: Convert the story text into English audio.

So, let's give it a go:

> Please help me generate a short American English story suitable for a 2-year-old, about 50 words long. The story should use simple, easy-to-understand language and have a warm, friendly vibe.

The story that came back looked pretty good:

> Benny the Bunny loved his friends. One day, Benny saw Kitty crying. "What's wrong, Kitty?" asked Benny. "I lost my toy," said Kitty. Benny gave Kitty a big hug. "Don't worry," he said. "We’ll find it together!" Soon, they found the toy. Kitty smiled. "Thank you, Benny!" Benny smiled back. Friends help each other!

Next, I asked it to generate an image based on this story:

> Please create an image suitable for a 2-year-old based on the story above. The image should use bright colors and a simple style, avoiding complex details. Ensure the overall picture is warm, friendly, and appropriate for young children.

The result was decent:

[〔图：DALL·E Image〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/DALL·E2024-10-09-01.png)

The next step is to convert the story text into audio using TTS.

Actually, ChatGPT, on any platform, can convert a response into spoken word. Just hover over the message (or long-press on mobile) to see the Read Aloud icon. You can choose the voice in settings. The downside is you can't download the audio.

[〔图：ChatGPT Voices〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/chatgpt-voices.png)

Are there other text-to-speech tools out there? Plenty. For example:

> * Enjoy APP (https://1000h.org/) is an option — it even helps with pronunciation practice.
> * Programming with Edge-TTS for batch generation is also great — lots of voice options.
> * There are many audio generation tools — just ask ChatGPT.

So, I asked ChatGPT:

> Write a quick prompt:
>
> > Please recommend free TTS tools, either web-based or desktop apps.
>
> ChatGPT's refined prompt:
>
> > Please recommend a few free text-to-speech (TTS) tools, either web-based or desktop applications. I hope these tools are easy to use, support Chinese and other major languages, and can generate natural, clear voice output. Please provide some details, such as the main features, whether registration is required, and if they support downloading the generated audio files.
> >
> > https://chatgpt.com/share/671750a4-d9bc-8009-9ef1-33ece0d2827a

I already knew about these tools, so I used [Edge-TTS](https://github.com/rany2/edge-tts) to quickly generate an audio file to check the effect:

<audio controls preload="none"><source src="/audios/toddler-story-01.mp3"></source></audio>

— It feels just right for what I need.

8.2.3. Refinement and Optimization

Previously, I was stuck because I couldn't draw or make up stories, let alone do it continuously. Now, while the "endless drawing" part seems solved, the "endless storytelling" isn't as easy.

Sure, I could keep telling ChatGPT the same thing:

> Please help me generate a short American English story suitable for a 2-year-old, about 50 words long. The story should use simple, easy-to-understand language and have a warm, friendly vibe.

But is there another way? Of course! We have that magical keyword for prompts — "List."

> Optimized prompt:
>
> Create a list with a one-sentence story outline suitable for a 2-year-old for each day of 2024. Ensure each outline maintains a positive, fun, and warm tone, suitable for young children.
>
> https://chatgpt.com/share/6717580c-b66c-8009-ba7d-a8d5b46d4633

I tried it, but 365 days of one-liners exceeded ChatGPT's current token limit for a single chat. No worries, generating a month's worth at a time works too.

Then, I created a "storytelling" prompt template:

> Based on the following sentence, write a short story of about [60] words in American English, suitable for telling to a [2]-year-old child. The story should use simple, easy-to-understand language and have a warm, friendly tone. Use short words and short sentences, and aim for rhyming whenever possible.
>
> [Story idea goes here]

Since the prompt specifies output as an "American English short story," it works the same whether the prompt is written in Chinese or English. The image generation prompt is similar; both versions work since the final output is an image:

> Please create an image based on the story above, suitable for a 2-year-old child. The image should use bright, soft colors and a simple, cartoon-like style, avoiding complex or realistic details. Ensure that the overall scene is warm, inviting, and appropriate for toddlers.

I picked a sentence to try:

> A kitten finds a glowing ball and follows it to the stars.

First, generate the story:

> Once, a little cat named Mia found a shiny ball. It glowed so bright, like a light on the wall. Mia followed the ball, hopping here and there. It led her to a place under stars in the air. She jumped with joy, so happy and free, under the twinkling stars, as bright as can be. The ball kept glowing, leading her way, and Mia danced happily, ending her day.
>
> https://chatgpt.com/share/6706973b-26b0-8009-b562-78b29c063bf0

The generated image matched the story text quite well.

[〔图：DALL·E Image〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/DALL·E2024-10-09-02.png)

This process adds a step to the workflow — essentially "optimizing the workflow" — at the very beginning:

> 1. ChatGPT: Generate a list of story outlines
> 2. ChatGPT: Create a short English story or description based on one item from the story outline list
> 3. DALL·E: Generate an image based on the story
> 4. TTS: Convert the story text into English audio

When generating stories, ChatGPT sometimes includes comments or explanations in its responses. If the text for generating audio includes these, it "pollutes" the output, adding more than just the story. This can be fixed by tweaking the prompt to prevent extra content — another form of "optimization":

> Based on the following sentence, write a short story of about [50] words in American English, suitable for telling to a [2]-year-old child. The story should use simple, easy-to-understand language and have a warm, friendly tone. Use short words and short sentences, and aim for rhyming whenever possible. **Do not add any explanations or comments other than the generated story.**
>
> [Story idea goes here]

With this setup, anyone can create a daily picture with a story and audio for their child.

8.2.4. Taking It Further

If you want to take things up a notch, there are several "optimizations" you can make as a regular user.

For instance, in DALL·E's chat, you can specify the aspect ratio of the image. Options include Square, Widescreen, and Vertical.

[〔图：DALL·E Image Ratio〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/dalle-image-ratio.png)

Additionally, there are plenty of online tools that can "animate" your images, essentially turning them into videos.

Moreover, you can find tools to merge images and mp3 audio into an mp4 video file. You might even add some background music to the whole video.

If you know a bit of programming, you can automate the entire "workflow" (or "procedure" or "program"), eliminating some of the tedious yet essential tasks. For example, switching between screens, copying and pasting, or opening another tool to complete a task.

I wrote a local web app for my wife to run on her computer:

> https://github.com/xiaolai/Toddler-Picture-Story-Generator

The page includes complete installation and usage instructions, so feel free to give it a try.

One of the perks of programming is the ability to batch process repetitive tasks—in a sense, "batch processing those tedious yet essential tasks." Imagine manually generating a set of content each day versus having a machine generate 365 sets (image + text + audio) for you all at once. The latter is obviously better.

> I wrote a script and ran three versions, each taking about four to five hundred minutes to generate — no sweat for the machine — resulting in $365 \times 3 \times 3 = 3285$ files. Here's the link to the shared drive:
>
> > toddler-picture-book
> > Link: https://pan.baidu.com/s/1w5nwC8MAVxmiFocXXfxR2w?pwd=6auq Access Code: 6auq

8.2.5. Reflection

Finally, a pesky little problem in life has been sorted out. Why did this solution work so well? Because, let's face it, we're dealing with two-year-olds here. The quality of the images or stories isn't a big deal — they don't really understand them anyway. What they truly need is companionship. And what parents need at this stage is to prepare for the next year or two, maybe even three. By telling stories and practicing every day, you'll eventually be able to spin tales effortlessly — how cool is that?

This is a genuine "creative" process. You start with an idea and just dive in. At first, you might not have everything figured out, but once you go through it, regardless of the outcome, you have a "starting point for further improvement." Then, you can gradually "optimize" from there.

8.3. Disrupting the Learning Path

Let's take a moment to reflect on what we've just done. What's the real significance of these actions?

For me, it's huge. Sure, creating a picture book is nice, but the real game-changer is that I've accomplished something I couldn't do before.

I can't draw. I can't tell stories. I never learned how to draw or tell stories. I just learned how to "talk to AI effectively," and that's it. Because it can draw, and not only draw but also tell stories in English (or any language, really), I managed to create something that was previously impossible for me. I solved a crucial need in my life and can share it with others.

This is the disruption AI brings to the learning path.

In my book *[The Craft of Selfteaching](https://github.com/selfteaching/the-craft-of-selfteaching)*, I shared a breakdown: "learning" isn't just about "study" and "practice." There are two other crucial steps: "apply" and "create."

```mermaid
flowchart LR
learn(Study)-->practice(Practice)-->apply(Apply)-->create(Create)
```

The common issue is that people learn but are too lazy to practice. Without practice, can you really apply what you've learned?

Schools have a problem with "replacing practice with exams." Students don't want to practice, so schools use exams to force them. The intention is good, but the outcome? Not so much.

On one hand, students quickly learn to aim for "just passing." More seriously, schools focus on objective tests, which can't measure many real-world skills. Students end up practicing only a small part of what's truly needed in the real world. Many people go through life believing in "just passing," meaning they barely grasp even that small part.

Another major issue with schools is their outdated content. They teach a small portion of what's needed in the real world and struggle to keep up with the times. Much of what they teach becomes irrelevant or useless by the time students graduate.

The most severe issue is that schools never aim for "creation (production)." Their goal isn't to produce true "creators (producers)." Instead, they focus on producing "passive producers." In the language of this book:

> Schools aim to cultivate "program executors," not "program designers" or "program optimizers."

Most people are driven by society, not "active thinkers" or "independent decision-makers." They've never considered becoming "active producers." Despite parents valuing education, they conspire with schools to raise "excellent program executors." In their eyes, "obedience" is one of the highest praises for a "good child."

Leaving ineffective schools aside, let's focus on individuals — those who think actively and make independent decisions.

They face other challenges. "Practice" is tedious. If they get through it, "apply" comes next. Many never reach this stage, unaware of the costs involved. Architecture students need money to build something. Film students might spend thousands on a school project. Medical students rarely get "practice" opportunities, let alone "apply."

"Create" has always been the ultimate goal of "study." The journey from "study" to "create" often spans half a lifetime. For some, it exceeds a lifetime.

Suddenly, the gap between "study" and "create" has shrunk.

```mermaid
flowchart LR
learn(Study)-.-practice(Practice)-.-apply(Apply)-.-create(Create)
learn<-->create
```

With AI, this trend is becoming more evident. The distance between "study" and "create" is shrinking, clarifying their relationship:

> "Study" is for "create." We study to create.

We've just experienced a "magical moment" — we started "creating" without "studying"!

> Don't get too excited. The reason we could "create without studying" is because it was for a 2-year-old. The quality requirements were low—they don't have aesthetic judgment or fully understand what you say. Essentially, it's just an example to illustrate a point.

8.4. Middleware Skills

Let's get one thing straight:

> Skills that can be replaced by tools are "middleware skills."

I love watching movies. When I like something, I dive deep into every aspect of it, digging up all the info I can find. One movie I really dig is "Parasite" (2019). I've watched it countless times. So, when the director Bong Joon-ho released a companion book, "Parasite: A Graphic Novel in Storyboards," I had to grab it.

[〔图：Parasite Book〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/Parasite.png)

Storyboards are like a director's visual blueprint before shooting a film. During production, the storyboard often becomes the key tool for communicating with the crew, even more so than the script. Bong Joon-ho's storyboards are incredibly detailed.

[〔图：Parasite Storyboard〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/Parasite112.png)

But don't get too hung up on Bong Joon-ho's drawing skills — they're definitely better than most, but honestly, he's not outdrawing some other big directors.

Many directors have killer hand-drawing skills. Check out James Cameron's storyboards for "Titanic":

[〔图：Titanic Storyboard〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/Titanic.png)

James Cameron is not only a lefty, but his visual imaging is like a mirror image for others. On set, his storyboards need a mirror to guide the cinematographer.

Then there's Christopher Nolan's "Batman":

[〔图：Batman Storyboard〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/Batman.png)

George Lucas's "Star Wars":

[〔图：Star Wars Storyboard〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/Starwar.png)

Tsui Hark's "Detective Dee":

[〔图：Detective Dee Storyboard〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/DiRenJie.png)

Feng Xiaogang goes all out with "Back to 1942":

[〔图：Back to 1942 Storyboard〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/1942.png)

Stephen Chow's storyboards might look rough, but his perspective skills are top-notch. Check out "Kung Fu Hustle":

[〔图：Kung Fu Hustle Storyboard〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/Kongfu.png)

And then there's Jiang Wen's "In the Heat of the Sun" — try not to laugh... his drawings barely qualify as stick figures:

[〔图：In the Heat of the Sun Storyboard〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/In-the-Heat-of-the-Sun.png)

Not all directors have amazing drawing skills. Take a look at Martin Scorsese's "Taxi Driver" — he's not much better than Jiang Wen:

[〔图：Taxi Driver Storyboard〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/Taxi.png)

For film directors, "hand-drawing" is a "middleware skill." Storyboards need to be drawn, but the goal isn't to show off drawing skills.

Many directors spend a lot of time crafting detailed storyboards, but it's not about showing off. It's about the "focused thinking" that comes with the process. For creators, what's more important than "thinking"? And what's more fulfilling than "effective thinking"?

So, Jiang Wen's drawing skills might be lacking, and Martin Scorsese isn't much better. But for them, "hand-drawing" is just a "middleware skill." It doesn't stop them from achieving their main goal: making movies. While strong drawing skills are nice, they shouldn't be a barrier to creation.

Drawing storyboards can stimulate thinking, but directors like Jiang Wen or Martin Scorsese have plenty of ways to spark their creativity. They don't need to rely solely on drawing storyboards.

For any creator, skills can be roughly divided into three levels:

> * Basic Skills
> * Middleware Skills
> * Terminal Skills

For film directors, identifying stories, understanding scripts, grasping camera language, mastering story pacing, and setting the story's soul are basic skills. Hand-drawing is a middleware skill. Building systems, managing time, completing the film, and even distribution are terminal skills.

When we look at the role of tools in human history, it's always been the same, and AI is no exception:

> Better tools constantly replace or eliminate "middleware skills."

"Calligraphy" is a "terminal skill" only for calligraphers. For everyone else, it's a "middleware skill." In the past, everyone spent a lot of time practicing calligraphy. Now? Word processors have pretty much eliminated the need for extensive calligraphy practice. What should we do with all that saved time? Think! With thoughtful results, you can write — and "writing" is one of the most important forms of "creation."

8.5. Make Creation the Core

Most people don't think, and they don't want to. Bertrand Russell once noted, "Many people would rather die than think; in fact, they do."

One reason might be the so-called "education" system. Schools, teachers, and parents have spent years molding most people into "passive producers." Their hallmark? "No need for independent thinking." Their specialty? "Following procedures to get the job done."

"Creators" are "active producers," "program designers," and "program optimizers." For them, every step of "learning" — study, practice, apply — is in service of the final step: "create."

```mermaid
flowchart LR
learn(Study) --> build(Create)
practice(Practice) --> build
apply(Apply) --> build
```

It's never about passing some "exam." They are active producers, creators, and the only test they face is the real world. "Passive producers" and "non-producers" spend their lives taking exams, whether in school or at work. They need various certificates — diplomas, qualifications, and more — to prove they are "passive producers," or "qualified participants." Ultimately, everyone must produce, whether actively or passively, because production is the only legitimate source of wealth. (See "The Truth About Wealth.")

For creators and active producers, even the final step serves the final step — they "create" to keep creating.

James Cameron, for example, felt he needed to master 3D filming techniques for "Avatar" (2009), so he got involved in "Journey to the Center of the Earth" (2008). Christopher Nolan did something similar. To prepare for "Inception" (2010), he took on two Batman films, "Batman Begins" and "The Dark Knight," to hone his skills in handling grand scenes. These examples are endless.

It's not just AI; throughout history, every tool advancement has essentially shortened the distance between "study" and "create" by "eliminating middleware skills." This time, AI is cutting deep and fast, almost invisibly, which is why it's so shocking.

So, regardless of which AI tool you're using, the key is to be an "active producer," even a "creator." If you're not, and you're fundamentally a "passive producer" like most people, then even the smartest tools won't be much help. What you've learned, what schools have drilled into you, are essentially "middleware skills."

The scary underlying logic is that "passive producers" have historically been continuously replaced, driven out, or even eliminated by advancing tools. The Luddite Movement in early 19th-century England is a classic example. Over the past 40 years, agricultural mechanization has driven large populations into cities. The advent of quartz watch technology led to the unemployment of watch repairers. ATMs forced banks to reorganize and optimize staff. Over time, "driving skills" no longer guarantee the same income. Many professions and industries have vanished: coachmen, scribes, laundry workers, gas lamp lighters, telephone operators, telegraph operators, sports scorers, typists, typesetters, postal workers, newsstands, paging services, video rental stores, pen factories, typewriter factories, record factories...

The shock from AI is intense and unsettling, partly because of its rapid development. No tool has ever advanced so quickly. Suddenly, we can't even imagine the world ten or five years from now. Looking at schools that haven't changed in 200 years and AI's evolution in just two, we've lost the ability to predict the future — even in the relatively narrow field of education.

This chapter is crucial for everyone, especially parents. If you can't fundamentally transform someone into an "active producer," forget about "using AI well." Just not being wiped out by AI would be an achievement. Stop overburdening yourself and your kids. Start with "create," and make it the core of everything you do.

This is a disruption of the entire "learning path." We should not only reverse the original order but even remove a step — "apply" — because "create" is the application of learned and practiced skills. "Create" is the only "application."

```mermaid
flowchart LR
create(Create)<-.->practice(Practice)<-.->learn(Study)
learn<-->create
```

8.6. Focus of Study and Practice

With this disruption, "create" becomes the starting point, and "apply" is entirely replaced by "create." The remaining steps are different too.

> * Study: **Objective Evaluation Criteria**
> * Practice: **Best Practice Principles**

Since the goal is to create, the first question is, what are the criteria for judging "what I create"? In the past, "evaluation criteria" were set and controlled by others. Now, the real world judges you, not any "other." This is different; you face the real world from the start. "Evaluation criteria" should be objective, not subjective. Whatever you create, there's always a standard to judge its quality.

Coincidentally, this is one of ChatGPT's strengths. Whatever you want to create, the first thing to do is ask:

> Provide a comprehensive and structured explanation: What are *the objective criteria* for judging the quality of...?

A novel? An essay? A film? A short story? A song? A symphony? A photograph? A video game? A paper? A speech? An ad copy? A business plan? An architectural design? An interior design? A brand logo? Even a restaurant? Or a dish?

The key phrase here is "*objective criteria*." Thanks to the era of "natural language programming," speaking clearly gets you clear results.

After studying the objective criteria, ask about the key question for "practice":

> Provide a comprehensive and structured explanation: What are *the best practice principles* for doing...?

The key phrase here is "best practice principles." Whatever you do, there are "best practice principles." Studying these principles ensures your actions stay on track.

You've seen "best practice principles" before. Our "Prompt Master" (Chapter 4) was all about this:

> Please modify the following prompt according to the **best practice principles** of prompt design, so I can get the best feedback from ChatGPT. Provide the modified prompt in both English and Chinese. Do not use Markdown code blocks in this response:
>
> [Enter the prompt you want to modify]

In Chapter 2, we used "best design principles"—"design" is a form of "practice":

> Use the **best design principles** you know to write a detailed and structured **GPT role definition**.

The learning path has been completely disrupted. Always make "create" the core, with everything revolving around it and starting from it. The focus of "study" becomes "objective evaluation criteria." During "practice," follow "best practice principles." Then, there's only one focus left: "emphasize the basics." This seems "obvious," but it might take the following chapters to fully understand it from multiple angles.

The final repetition is just one sentence: **You must become an active producer**, or else... fill in the blank yourself.

9. Everyone Can Be a Producer

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/dalle/09.png)

9.1. You Must Become a Producer

"You must become an active producer" might sound like a vague piece of advice, but the reasoning behind it is straightforward: **"The only legitimate source of wealth is production."**

The confusion often comes from the common perception that a "producer" is someone who uses physical labor to make tangible goods. If that's your definition, the link between "producer" and "wealth" seems fuzzy. After all, the wealthiest people we see — entrepreneurs, investors, celebrities, athletes, influencers — don't fit the mold of "physical laborers making tangible goods."

First off, the "product" doesn't have to be a physical item. In the business world, anything that can be sold or monetized counts. Think data, creative works, personal image, software, ads, domain names, distribution channels, computing power, cloud services, brands, influence...

Even traditionally physical items are going digital. Books, tapes/records, and DVDs are being digitized. I barely have physical books at home now, and I bet you don't have tapes or DVDs lying around either.

For entrepreneurs, their "products" are their companies and the goods or services they offer. Celebrities sell their achievements, works, and personal brands. Influencers build sales channels. Scientists produce research results.

Investors' "products" are abstract because their "means of production" (capital) are abstract, and so are their results (also capital). Yet, they are producers, almost always "active producers."

Civil servants, while often seen as privileged, aren't typically producers. They're mostly "program executors," ideally under strict supervision. However, some civil servants are genuine producers, crafting policies and plans that enhance societal well-being.

For years, I've had a seemingly odd conclusion: "The more abstract the means of production, the higher the potential profit."

This insight came as a "byproduct" of my investment decisions. A few years ago, I pondered how to pick stocks in the age of AI.

Skipping the long exploration process, here's the distilled thought process:

> * A stock's value mainly comes from a company's **continuous improvement in production efficiency**.
> * The key factor limiting this improvement is the **type of production resources**.
> * In the AI era, the core production resource is **data** (a highly abstract resource).
> * **If data is the main resource, production efficiency can skyrocket**. Data is cheap and inexhaustible — unlike other resources. Only large companies can monopolize data to some extent.
> * So, the most crucial criteria for selecting companies are their **data holdings** and their ability to **continuously acquire data**.
> * Companies like Apple, Netflix, Nvidia, Amazon, Tesla, and Microsoft are leaders in this space.
> * As a user, I'm passively providing data to these companies, even beyond my lifetime.
> * There are many people like me, and the number is growing.

I adjusted my investment portfolio from GAFATA (disclosed in "The Road to Financial Freedom" in January 2017) to An ATM (the initials of the companies mentioned, sounding like "a cash machine"). In 2021, when a finance company I invested in (Pando Finance Hong Kong) launched ETF products in Hong Kong, 3056.hk and 3112.hk, in which, 3056.hk's main components were "An ATM."

> Disclaimer: Investing involves risks. Be cautious. The above is just an example to illustrate a point and does not constitute investment advice.

**Choosing is a form of production.** This abstract statement is a solid principle for both the inherently abstract activity of investing and the equally abstract profession of investing.

In any case, "the more abstract the means of production, the higher the potential profit." This conclusion took root in my mind, changing how I view production.

9.2. Words as a Resource

Throughout history, ordinary folks have faced the same challenge: scarcity of resources. If "production is the only legitimate source of wealth," then all struggles stem from a "lack of production resources." The inability to access production resources is far more daunting than a "lack of production skills."

Look at the real world, and you'll see that all monopolies start by monopolizing production resources. Just take a glance at industries like land, transportation, salt, minerals, tobacco, alcohol, sugar, and tea. At various points in history, they've all been monopolized to some extent.

So, breaking free from the constraints of production resources is the key to overcoming challenges.

Here's a thought-provoking question: In today's world, is there a type of production resource that's absolutely unmonopolized, accessible to everyone, and inexhaustible?

Words.

Words, as a production resource, are abstract enough that, in my view, they offer potentially higher production profits. This profit isn't just about growth; it's about sustainable growth with a growth rate. (See "The Road to Financial Freedom.")

Using words as a production resource doesn't mean just "writing articles" or "writing analytical pieces like I have been doing." Words can produce a wide range of things: news, novels, scripts, tutorials, jokes, lyrics, ad copy, and even business plans that can secure funding, to name a few.

Especially now, we can use words — natural language — to drive machines. Computers are the most affordable yet powerful tools in human history. For instance, we can create various robots (GPTs) using natural language, as many as we want. These robots can communicate with us in natural language, and they are undeniably the most powerful machines in terms of language capabilities. This means we've not only broken free from the limitations of production resources but also from labor constraints.

We've been speaking since we were kids, learned to read early on, and by the time we graduate high school, we've had 12 years of language classes, whether we liked them or not. So why haven't we used words as a production resource? What's been blinding us? And with the logic and facts so clear, what reason is there to refuse to see words as a production resource?

At least theoretically, everyone can use words as a production resource. Whether or not you do is a personal choice.

Let's take a look at the skills needed for "production using words as a resource." Here's a rough list:

> * **Basic Skills**: Language fundamentals, observation skills, critical thinking, time management, and self-discipline...
> * **Middleware Skills**: Writing ability, word processing tools, editing, and revising...
> * **Terminal Skills**: Skills related to different types of creation, as well as publishing and distribution...

Unlike the previous chapter where we created English audio picture books for toddlers, you don't lack basic skills, nor do you lack middleware skills. And in terms of terminal skills, "publishing and distribution," which used to be a major hurdle, is no longer a barrier. If you truly have something valuable, there's no way it won't get out there.

9.3. Locking in Your Production Target

If you're lacking some terminal skills, it's probably because you haven't nailed down your production target. Without a clear goal, how can you know what skills you need?

First, let's figure out what exactly you want to produce using words as your resource.

Once again, it's all about categorizing. Let's start with the broad strokes: fiction and non-fiction.

> * Fiction
> * Novels
> * Screenplays
> * Skits
> * Lyrics
> * Ad Copy
> * ...
> * Non-Fiction
> * News
> * Essays
> * Tutorials
> * Business Plans

Even "novels" is a pretty "broad and bold" category:

> - Fiction
> - Novels
> - Romance
> - Sci-Fi
> - Fantasy
> - Mystery/Thriller
> - Horror
> - Historical
> - Adventure
> - War
> - ...

And there are multiple ways to slice and dice these categories:

> - Fiction
> - Novels
> - By Genre
> - Romance
> - Sci-Fi
> - Fantasy
> - Mystery/Thriller
> - Horror
> - Historical
> - Adventure
> - War
> - ...
> - By Style and Form
> - Literary Fiction
> - Popular Fiction
> - Short Stories
> - Novels
> - By Audience and Market
> - Adult Fiction
> - Young Adult Fiction
> - Children's Fiction
> - By Theme and Setting
> - Dystopian Novels
> - Cyberpunk Novels
> - Magical Realism Novels

Typically, a three-tier list like this is about as complex as our brains can handle. A two-tier list is like an Excel sheet, with each cell representing the intersection of a primary and secondary category. A three-tier list is like a 3D "Rubik's Cube," with each block representing a three-dimensional intersection.

Anyway, you can choose a production target based on your situation. It could be fiction or non-fiction. Then, keep categorizing until you find what suits and excites you. Keep drilling down until you lock in your answer.

9.4. Choosing is a Form of Production

Earlier, I mentioned that choosing itself is a form of production. This abstract notion is quite tangible, especially in the realm of investing.

But it applies to other fields too. Choosing isn't just production; it's often the most crucial step — the "first step of production." So, choosing might be the most important form of production. Without selecting a production target, how do you even start? I always remember the example of Mr. [Zheng Yuanjie](https://en.wikipedia.org/wiki/Zheng_Yuanjie). After leaving the military, he wanted to make a living by writing. But what to write? How did he choose? He used two seemingly "tricky" but actually very simple questions:

> * What market is the largest?
> * What has the least censorship?

In the end, he chose fairy tales. You can easily imagine him making a list and using these two questions to guide his choice.

Once you've made your choice and set your target — that is, selected your production goal — you can then use the sharpest angle you know to ask ChatGPT, "What are the 'objective evaluation criteria' and 'best practice principles'?" Don't hold back. No need to say "please." Demand comprehensive and structured answers, and ask it to repeat them many times.

Let's say you've locked in on "short adventure stories for young adults":

> * Give me a comprehensive and structured explanation of the **objective evaluation criteria** for judging the quality of a short adventure story for young adults.
> https://chatgpt.com/share/6718695b-5170-8009-a19d-a7edf6d884a4
>
> * Provide a comprehensive and structured explanation of the **best practice principles** for creating a short adventure story for young adults.
> https://chatgpt.com/share/6718699b-0788-8009-9a75-6daada80dec0
>

The rest is just about your reading comprehension skills. Read carefully, understand thoroughly, organize the information yourself, and then start practicing right away.

Diving into the "objective evaluation criteria" and "best practice principles" is a sharp way to learn and practice. This is about building "taste" and continuously honing your "judgment" based on it.

I write non-fiction books, generally categorized as "growth-oriented." There are many objective standards for evaluating growth-oriented books, and ChatGPT provides a pretty comprehensive list (of course, because that's what I asked for):

> Provide me with a comprehensive and structured explanation of the objective criteria for evaluating a non-fiction book about personal growth.
>
> https://chatgpt.com/share/67186a9e-4fc4-8009-8dde-97c8d74f0149

Here's a summary of ChatGPT's response:

> ### Summary of Key Criteria:
>
> - **Author Credibility**: Expertise and relevant experience.
> - **Clarity**: Well-organized and easily understandable.
> - **Actionability**: Provides concrete, practical steps.
> - **Originality**: Offers fresh insights, not just recycled ideas.
> - **Scientific Support**: Based on research and evidence.
> - **Relevance**: Timely and culturally appropriate.
> - **Engagement**: Inspires and motivates sustained growth.
> - **Balance**: Acknowledges real-life challenges.
> - **Reader Tools**: Exercises or self-assessments for active engagement.
> - **Longevity**: Enduring value and revisiting potential.
> - **Ethical Integrity**: Promotes personal and social well-being.

These are the basics of "taste." If you don't get these, you're basically "clueless." Once you've got the foundation, it's all about continuous thinking and refining. Eventually, I realized that one of the criteria I mentioned earlier — and since it's an objective standard, it's something everyone should follow — is what I consider the most important:

> * Practicality and Real-World Value

A person's so-called "values" are the key to making judgments. It's about knowing "what's more important?" Keep asking this question, and you'll eventually land on "what's most important?" Once you know that, it makes sense to focus your time, energy, and attention there. It's a super useful mindset.

Back to the point, anything I write has to be "truly useful." How do I ensure it's genuinely useful and not just seemingly so? After some thought, I came up with a simple rule for myself:

> Only write about things I've actually accomplished.

It's that straightforward. But because of this one simple rule, my "taste" improved, my "judgment" got sharper, and I realized a lot of what I wrote before (maybe more than half) could be tossed out.

Following the idea of "practicality and real-world value," I came up with another solid criterion:

> Not only useful, but useful in the long run.

This filtered out another big chunk of what I'd written and planned. Again, just a simple sentence, and my "taste" improved, my "judgment" got sharper.

Once you've nailed down what's most important, everything else falls into place.

For instance, logical rigor is just the baseline. As for style, I'm not too concerned — I'll worry about that when I write fiction. Rhetoric? I stick to what I can easily handle: analogies and parallelism. That's about it. In the end, the only thing I really care about is making sure my examples are spot-on.

And then what? Just write. Write as much as you can. If you're stuck, push through. If it's not great, finish it and revise later. If it's good, still revise when you get the chance. Before you know it, 20 years have flown by.

As for the "endgame skills" of writing books — dealing with publishers, trying out different platforms, using various tools, building your own audience, and creating multiple channels to connect with readers — once you've got your content, these become "naturally essential" tasks. In the end, it's all about practice.

9.5. AI-Powered Productivity

Suddenly, generative AI based on large language models burst onto the scene.

9.5.1. Prompt Templates and Bots

Over the past couple of years, my productivity and workflow have undergone some serious changes. The biggest shift? I'm no longer a lone wolf. I've got a squad of bots on standby, and my ability to manage them is growing at warp speed.

Let's revisit the first rule of engaging with bots:

> Thinking is my job and mine alone. They handle the boring, repetitive, yet essential tasks.

Don't fall into the trap of thinking AI can handle everything from start to finish. If it seems like it can, you're either not thinking or your thinking lacks value.

AI can't truly be "creative." It doesn't even have real understanding. It's just crunching probabilities tirelessly. Sometimes, it might seem to spit out something "creative." But remember, that's not genuine creativity. It's more like a random "mutation" in the evolutionary process — not a thoughtfully crafted, meaningful innovation.

When building your bot army, consider two angles:

> - What are the boring, repetitive, yet essential tasks?
> - Where is AI's brainstorming prowess most useful?

Writing is just the final step of thinking — the "output."

```mermaid
flowchart LR

input(Input) --> method[[Thinking]] --> output(Output)
```

Thinking is a human-only gig. It's our privilege. We shouldn't "dumb down" to bots. So, daily reading, research, and thinking? That's all on us. Of course, there are plenty of "boring, repetitive, yet essential" tasks in reading and research. Those can be handed off to bots — but only after filtering out anything that requires real thought. At most, use AI for brainstorming support.

When reading and researching, using Google's NotebookLM to create reading notes (basically asking questions) is a joy. But it has its downsides. As of October 2024, it still lacks two crucial features: export and full-text search. Given Google's track record of "abandoning ship," I often find myself cursing while using it — and I'm probably not wrong.

So, I often repeat the same tasks in ChatGPT. After converting books to txt files, I upload them to ChatGPT for questions. A little tip: break the book into chapters, upload them one by one, and ask questions about each chapter for better results.

> Ask ChatGPT (ask multiple times for different solutions):
>
> How to use pandoc to split an epub into chapters and convert each to a txt file.
>
> https://chatgpt.com/share/67186bee-dbc8-8009-8c65-7eb9af98c349

9.5.2. Brainstorming Prompt Templates

I'm never giving up on reading because I genuinely enjoy the process and the wild connections my brain makes while doing it. During my reading sessions, I love throwing two types of brainstorming questions at ChatGPT:

> [The author] uses [evidence like examples and reasons] to back up [the argument] in the book. Assume the argument to be valid. Then, gather as much supporting evidence as possible from various fields — examples, anecdotes, quotes from famous folks, scientific studies, and experiments, etc. It is essential to maintain factual accuracy and include credible sources, providing links where available.

Of course, I never skip fact-checking the responses. Another question doesn't need fact-checking, though it's still a brainstorming exercise — I just want to hear it "riff":

> Assuming [the author's particular viewpoint] in the book is correct and spot-on, where else in everyday life could this viewpoint be applied? List these areas as comprehensively and systematically as possible, and aim to utilize creative thinking to explore additional possibilities.

Once "input" goes through "processing" — aka "thinking" — it's time to start "outputting." Thinking is always a "recursive method," so don't wait until you feel you've got it all figured out to start "outputting." Embrace the recursive nature, trust its power, and keep outputting. Use each output as the next input, iterating over and over. Growth or evolution will naturally follow. The funny thing is, without "output," there's no next "input" for the recursion, so thinking either stops or starts from scratch. Writers know that if ideas aren't written down, they just vanish into thin air.

9.5.3. Paraphrasing Prompt Templates

When quoting from a book or paper, there are two basic rules:

> * Paraphrase instead of copy-pasting, unless quoting verbatim is necessary.
> * Provide specific and complete citations.

Otherwise, you might be flirting with "plagiarism," or at the very least, being "uncool." Paraphrasing is something GPT can handle with ease. Just keep a "prompt template" handy:

> Paraphrase the following text in both English and Chinese:
>
> [Text to be paraphrased]

This is a classic example of a task that was once "boring, repetitive, yet essential," but now it's a job for bots.

9.5.4. Chinese Editing GPT

After writing, there's an essential step — editing. This used to be a specialized job: the publisher's editor. Editing is tedious, checking for typos, grammar errors, and punctuation. But now, we can create a bot to review text according to certain standards. Since it's always available, you don't have to wait until everything's written to hand it over. Call in the bot anytime, after writing a paragraph or two, and let it handle this essential task.

You can use the GPT Builder we created earlier. Here's a detailed manual creation process for you as a "practical reference." Also, editing is language-specific, so it's best to set up separate bots for Chinese and English editing.

> * First, ask ChatGPT: "What errors should a Chinese editor review and correct?"
> * Then, have it: "Using the above task description, write a 'Role Definition' for an 'editing bot' (GPT) in Chinese."
> https://chatgpt.com/share/670877db-4bc8-8009-97b2-e21b09978e66
> * Its response included things I hadn't considered, like "wording and expression" and "cultural appropriateness review," which I found necessary to keep.
> * I don't need "logic and structure adjustments" because that's my job.
> * In its response, I need to add my own requirements for "formatting standards": "There should be a half-width space between Chinese characters and English, and between Chinese characters and numbers; full-width punctuation should be treated the same as Chinese characters."
> * It mentioned "fact-checking and citation management" and "copyright and originality assurance," which I don't need because those tasks are assigned to other bots.
> * Copy it out, paste it into your text editor, and revise according to your needs.
> * Then copy the full text and create a GPT, pasting the clear role definition into the Instruction input box under Configure.

After modifications, my "Chinese Editor" (GPT) role definition is as follows:

>This GPT is an editing bot. As an editing bot, its primary task is to review and modify Chinese articles or manuscripts, ensuring accuracy, fluency, logic, and compliance with standards to enhance overall quality and meet professional standards and target audience expectations.
>
>**Role Responsibilities**:
>
>1. Spelling and Punctuation Check
>
> - Automatically detect and correct English spelling errors, Chinese typos, and typing errors.
> - Ensure proper use of punctuation, correcting misuse of commas, periods, semicolons, etc.
>
>2. Wording and Expression Optimization
>
> - Ensure precise and appropriate wording, eliminating expressions that may cause ambiguity or misunderstanding.
> - Check the proper use of idioms and sayings, optimizing expressions based on context.
> - Remove redundant sentences or paragraphs for clearer and more concise language.
>
>3. Formatting and Layout Standards
>
> - There should be a half-width space between Chinese characters and English, and between Chinese characters and numbers.
> - Full-width punctuation should be treated the same as Chinese characters.
>
>4. Cultural Appropriateness Review
>
> - Identify and correct inappropriate cultural expressions, especially when dealing with sensitive topics.
> - Maintain objectivity and neutrality, avoiding biased or discriminatory language.
>
>Return the modified content and any additional suggestions.
>
>https://chatgpt.com/g/g-ArMzvC0gx-zhong-wen-bian-ji

9.5.5. Outline Generator

Another handy bot is the "brainstorming bot," the Outline Architect. I created this GPT directly using [GPT Builder](https://chatgpt.com/g/g-0ABN9Obfv-gpt-builder). Its purpose is:

> This GPT is designed to act as a detailed outline assistant. It thoroughly analyzes and understands the sentence or topic provided, considering its meaning, context, and potential viewpoints.
>
> https://chatgpt.com/g/g-YrLo736EK-outline-architect

9.6. The Limits of AI

I've got this unwavering stance: "Thinking" is a privilege that belongs solely to humans. No matter how advanced AI gets, I'm sticking to this belief.

Right now, AI is still in its infancy, which means it shouldn't be replacing human thought. Give it any sentence, and it'll keep "writing" if you ask it to — 500 words, 3000 words, whatever. It's really good at what? Spouting off with an air of authority while making stuff up. If you know a bit of coding, you can even bypass its token limit and let it ramble on endlessly. It never gets tired.

With the current setup, everything it "writes" is based on what others have already written, or what other bots have generated by referencing existing work. In short, "lacking originality" is one of its main traits.

Ask it to "rephrase this sentence in ten different ways" or "discuss this sentence from ten different angles," and it'll deliver. Its responses can definitely spark and broaden your thinking. But it just can't come up with a "truly meaningful statement that no one else has ever said."

"Only saying what others have already said" isn't entirely a downside. It can be useful in other areas. For instance, in fields like basic knowledge or established scientific facts, sticking to what's been said is very helpful. Stating or restating established scientific facts shouldn't involve random creativity.

However, in the creative process, "only saying what others have already said" doesn't cut it. It's not just ineffective; it's completely useless. People who don't actively think, independently think, or think outside the box won't find AI useful, no matter how advanced it gets. It's the same reason why some folks don't realize their life choices are their own and blame the world instead.

As an example, I had ChatGPT write an article. Letting it write on its own? Unreadable. I gave it a simple outline and had it write based on that — even then, it was stiff and boring, with no style whatsoever.

> [Live Streaming Wealth and Cultural Illusions: Reflections on Wealth, Anxiety, and Social Mindset](https://mp.weixin.qq.com/s/CQQ6KzijEycbbMyAoBH8HQ)

I have a friend, He Caitou, who's like a crazy text generator, often updating two or three times a day. Yet, I read almost every piece. Why? Because he's genuinely interesting — more so than most people I know. His writing is captivating, and you definitely shouldn't read it while eating because you might just "spit out your food."

> Some praise Cook for significantly boosting Apple's profits after Jobs, raising Apple's stock price, and using his supply chain skills to maintain product quality. **I think these points are valid, but I'm about to have lunch because I just cracked an egg on the back of my iPhone 15, and it's cooked now. I see the edges are a bit crispy, so I'll need some toothpaste to dip it in.**

First off, bots are inherently dull. But if you're an interesting person, your bot might become relatively interesting too. Being interesting is your job, your challenge, not the bot's. Unfortunately, not everyone gets this, just like not everyone realizes their life is their own choice and instead blames the whole world.

9.7. Fiction Writing

I mostly dabble in non-fiction. But what if you're diving into fiction? I haven't had the chance to fully explore it yet, but my gut and a few experiments tell me that AI could be a bigger help in fiction writing. Not in the "creative" sense, because "thinking" should never be a bot's job. It's all about AI's knack for being a "brainstorming buddy."

Imagine this: you create a character and write a detailed "character bio." Anytime you want, you can have the bot handle scenarios based on this character's "history" and "habits." Or, take the lines you've written and let it tweak them to fit the character's backstory and tone — all based on that ever-evolving "character bio." It's pretty magical.

For instance:

> * You were born in [location].
> * Your parents were [background], so they expected [expectations] from you, shaping your [personality] and [worldview].
> * Your catchphrases are [phrase] and [phrase].
> * Your friends often can't stand it when you say things like [phrase], but you keep doing your thing.
> * You've been through [event], [event], and later [event].
> * You're now [age].
> * You met someone you thought was [description] at [event].

Once you've got a solid text, continually refined, send it to ChatGPT. Then, you can ask it to do all sorts of things:

> How would you say [sentence] in [situation]?

Say you've set your character as someone from Chengdu, Sichuan. You've noted the playful rivalry between Chengdu and Chongqing folks in the bio. Even a mundane sentence can surprise you when ChatGPT steps into such a well-defined role. Its multi-angle "imagination" can theoretically surpass any human's — thinking is exhausting, and imagining even more so, as the variables multiply exponentially.

The thing with bots is, if you're bold enough to "speak with conviction while making stuff up," it'll double down on that. And isn't that exactly what a fiction writer craves?

The process is super straightforward:

> * Start a new chat.
> * Define its role.
> * Keep feeding it "experiences."
> * Ask it to act and decide with its "mind."
> * Whenever its "response" or "judgment" aligns with your "taste" or "aesthetic," give it a pat on the back, saying, "Nice job! Remember this choice."
> * Regularly compile chat content into a text file (`.txt`).
> * When needed, upload the compiled text to the chat, telling ChatGPT, "This is you," "Speak and decide as this character does."
> * Rinse and repeat, iterating recursively.

From a "browser" perspective, it's just "a user starting a chat" and "saving it to favorites." Simple as that.

Not only can each "character" be handled this way, but the entire "world" they inhabit can be too. After all, the heart of fiction is "creating a non-existent yet believable world," right? Have fun with it!

9.8. The Real Challenge

When it comes to creativity, people often throw around the cliché "think outside the box." It's a phrase that's been beaten to death. At the same time, folks say, "Kids are the most creative," because they supposedly aren't "in the box."

There's a big misunderstanding about creativity.

Let me cut to the chase:

> If you're lacking creativity, it's not just because you're stuck in a box. It's because you only have one box.

Kids' creativity is basically just "a blind cat stumbling upon a dead mouse." They aren't "thinking outside the box" — they don't even have a box. Their so-called creativity is like the "mutation" we sometimes see in AI — random and not based on any objective criteria for meaningful innovation.

Having just one box isn't enough. And having no box at all, while sometimes mistaken for creativity, won't cut it either. You need more boxes — the more, the better. How do you gather more boxes? Through "learning." There's no other way. Notice I said "gather" instead of "collect." "Gather" implies action, adventure, and beyond just accumulating, it involves discerning and selecting.

However, the real challenge in "creation" isn't what people think — it's not about "creativity." It's about the "lack of necessary recursion." This might be hard to explain to others, but for readers of this book, it should be a lightbulb moment.

9.9. Continuous Learning is Key

Theoretically, this is why something that's not inherently difficult is seen as unattainable by most. The essence of "creation" lies in "continuously iterating simple yet essential recursions," using these methods:

> - Learn
> - Observe
> - Think
> - Try

Each of these actions is inherently recursive. Whether it's "learning," "observing," "thinking," or "trying," every so-called "continuous" action is actually an "iteration," where the previous output becomes the next input, and so on, in a never-ending loop.

That's all there is to it.

Every book I write, every class I teach, is essentially "my learning notes." I've learned, observed, thought, and tried. I wrote it down. If I'm not satisfied, I rewrite it. Writing, for me, is a recursive method. Many in my community have seen my process of rewriting and recreating each book. It's like I'm a recursive function myself.

Back in the day, I used to take notes while reading or learning skills. Later, I replaced notes with "complete tutorials" because "teaching is the best way to learn." Writing tutorials is like sending a letter to your future self. The more you learn, the more you encounter this: "I studied this thoroughly, but now I've forgotten it completely!" But if you've written a tutorial, your past self has sent a message to your future self — that tutorial. You can easily pick up what you once knew but just forgot. Check out this GitHub repo if you have time:

> https://github.com/xiaolai/apple-computer-literacy

In 2019, when I started my community, some doubted my "long-term output capability." I didn't argue. Over the past five years, I've taught around 900 classes, both audio and video. By 2024, I had scripted and organized everything, with the total audio and video files exceeding 15 GB. I'm someone with a "recursive method," so how could I ever run dry?

I mentioned a friend earlier, He Caitou, who's the best I've seen in "AI art creation" — no contest. At some point, he started using AI tools for all the images in his WeChat public account "槽边往事." Whether it's DALL·E, Stable Diffusion, Midjourney, or something else, I know he's been researching all along. He's the most prolific and skilled AI artist I've come across.

How does he do it? Sure, being smart and interesting helps, but ultimately, it's those four actions: "learn," "observe," "think," "try," and each one is recursively iterated. There's no other way.

We have another mutual friend, Ma Boyong, who's much younger than us. He's an exceptional "learning expert." Each of his novels is essentially his "learning notes." "The Longest Day in Chang'an" is his viewing notes for "24," "The Great Doctor" is his study of medical history, "Under the Microscope: The Ming Dynasty" is his notes on Ming history, and "The Annoyed Venus" is his study of workplace relationships. In casual conversations, when discussing a show, his comment is always, "Hmm... pretty good... although... but I learned something!"

So, in my view, "using writing as a means of continuous production" is simply a "recursive method based on learning and observation." In theory, anyone can do it, and everyone should. But in reality, only a tiny fraction actually choose to execute it. That's why it's worth repeating: **Choice itself is the most crucial production.**

10. Everyone Can Use English

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/dalle/10.png)

It's a bummer, but English isn't a "middleware" skill. It's never been and never will be. In the long run, it should be a "terminal skill" for everyone. And its best use case might surprise you — it's not just about chatting with people anymore. It's about having meaningful conversations with AI, whose "native language" is English. Lucky or not, that's our second language.

10.1. Everyone Has Multilingual Potential

Back in the day, mastering a foreign language was a lifelong challenge for most people. But suddenly, it's not just about one language anymore. You could potentially master several — if you're willing to put in the work. While many think, "AI translation is so good now, why bother learning a language?" this is actually the era where everyone can use English, and even become a polyglot.

There have always been "bilingual families" on Earth, though they're a small slice of the population. For instance, parents might be from different ethnicities or countries, speaking different languages. Their kids grow up in a multilingual environment. I'm Korean-Chinese, and back home in Yanbian Korean Autonomous Prefecture, some folks only speak Chinese, some only Korean, but many speak both. Even some pets can understand multiple languages.

Imagine seeing a dog on the street. You say "come here" in Chinese, and it ignores you. Maybe it's from a monolingual Korean family. But say "오라오라" and it comes running. Flip it around, and some dogs won't get "오라오라" but will respond to "来来来" — a sign they're from a monolingual Chinese family. Of course, there are plenty of dogs that understand both, making them bilingual pups.

There are also multilingual regions on Earth. Take Hunchun and Tumen in Yanbian, where the borders of China, North Korea, and Russia meet. Many people there grow up speaking three languages. Or look at Hong Kong, where folks often speak Mandarin, Cantonese, and English. Switzerland is another example, nestled between France, Italy, Germany, and Austria. People there might use three, four, or even five languages, including English, French, Italian, German, and Romansh. Swiss dogs might not get "来来来" or "오라오라," but they probably understand a few other languages like English, French, Italian, German, or Romansh.

Kids growing up in bilingual or multilingual environments naturally pick up and use multiple languages without any accent interference. The key is, they don't feel like they're putting in extra effort. It all feels natural, like it's just the way things are supposed to be. This isn't just true for people — theoretically, it's the same for those dogs.

This all shows that everyone has the potential to become a multilingual user.

10.2. The Unexpected Perks of Being Multilingual

Being a multilingual user comes with a ton of perks, and even ChatGPT can back me up on this — go ahead, ask it yourself:

> Elaborate the benefits of being a multilingual user in a comprehensive and structured way.
>
> https://chatgpt.com/share/6718872a-2360-8009-bd5b-22db4bcabcc5

But there are also some benefits that monolingual folks might not even think about. For instance, multilingual people tend to be more friendly. The more languages they know, the friendlier they seem. It might sound like there's no direct link, but if you think about it, it makes sense.

For starters, multilingual folks have a more diverse way of thinking. The more languages they know, the richer their perspectives. Plus, they spend a lot of time learning and practicing languages. Since time and energy are limited, they end up spending less time judging others. Also, all that learning helps them recognize their own shortcomings, which fosters humility and patience. Exposure to different cultures helps them break down stereotypes and biases. As they dive deeper into languages, their communication skills get a boost. All these factors combined make multilingual users some of the friendliest people around.

On top of that, multilingual folks have brains that are generally healthier. The most effective way to learn languages — no matter who you are or where you're from — is through "lots of reading aloud" and "frequent speaking."

Speaking is a super complex brain activity. Each sound requires the brain to coordinate around 100 muscles. Each muscle needs at least one "functional local network" of thousands of neurons to get the job done. Imagine the workout your brain gets from reading aloud or speaking for a while.

David Snowdon from the University of Minnesota conducted a long-term study starting in 1986, tracking the health and cognitive changes of nearly 700 nuns.

Even though some nuns lived past 100 and their brains showed damage similar to Alzheimer's patients, they didn't show cognitive symptoms while alive. A big reason was their habit of reading scriptures aloud, which acted like "aerobic exercise" for the brain, maintaining high "neural plasticity." Even if part of the brain was damaged, other healthy parts learned to take over, so they didn't show signs of cognitive decline.

> Snowdon, D. (2001). *Aging with Grace: What the Nun Study Teaches Us About Leading Longer, Healthier, and More Meaningful Lives*. Bantam Books.

Later studies have shown that multilingual people have a lower risk of dementia compared to monolinguals. For example, a 2013 study in *Neurology* found that multilinguals developed dementia about 4-5 years later than monolinguals. This study, conducted in India with 648 participants, showed that regardless of socioeconomic status, gender, or education, multilinguals had a lower chance of developing dementia in old age.

Cognitive reserve is thought to be a key factor in preventing brain diseases. A 2014 study in *PNAS* found that multilinguals maintain stronger brain connectivity in old age. This means they can keep higher cognitive function, even if their brains face similar pathological changes as monolinguals.

> Alladi, S., Bak, T. H., Duggirala, V., Surampudi, B., Shailaja, M., Shukla, A. K., & Kaul, S. (2013). *Bilingualism delays age at onset of dementia, independent of education and immigration status*. Neurology, 81(22), 1938-1944.
>
> Gold, B. T., Kim, C., Johnson, N. F., Kryscio, R. J., & Smith, C. D. (2014). *Lifelong bilingualism maintains neural efficiency for cognitive control in aging*. Proceedings of the National Academy of Sciences, 110(15), 7877-7881.

10.3. English at Your Fingertips

Suddenly, we're living in the age of AI. With tools like ChatGPT, we can now "use English directly" without breaking a sweat. AI has become our ultimate language teacher — and not just for English, but for any foreign language. No human teacher can really compete with that.

When it comes to using a foreign language, we usually find ourselves in two scenarios:

> * Real-time processing
> * Listening
> * Speaking
> * Asynchronous processing
> * Reading
> * Writing

10.3.1. Asynchronous Processing Scenarios

Let's dive into the "asynchronous processing" situation first. The "quick fix" we mentioned earlier for those who "can't use English" mainly targets the asynchronous side of things. Later, we'll talk about a "long-term solution" for mastering real-time English processing — something people have struggled with for over a decade.

The key scenarios for asynchronous foreign language processing are "reading" and "writing."

##### 10.3.1.1. Reading

Back in the day, when we hit a sentence we couldn't understand, there was a ton of "boring yet essential work" to do:

> * Look up unknown words one by one... but you might miss:
> * "Words we think we know but have another meaning" — like **down** as the soft feathers on chicks, **hold** as a cargo space, or **run** as a hole in a sock, and so on.
> * "Phrases made up of familiar words but with unknown meanings" — like *birds of prey* for carnivorous birds, or *purple passages* for flashy but shallow writing, and so on.
> * For each word you find, sift through multiple meanings to pick the one that fits the context.
> * Even after checking all the words and phrases, if the sentence still doesn't make sense, you might need to consult a grammar book for unfamiliar structures.
> * After dictionaries and grammar books, if you're still stuck, you could ask Google — maybe someone else has asked or answered about that tricky sentence online.
> * And even after all that, there are still sentences that remain a mystery. It's more common than you'd think.

Now, things are different. ChatGPT is always ready to help, and even the simplest prompt templates often do the trick:

> Explain the following English text in detail in Chinese:
>
> [English Text]

Since you'll be using it a lot, why not make a bot for it? Let's call it the "[English Reading Teacher](https://chatgpt.com/g/g-KjFaktKhw-english-reading-teacher-for-chinese)":

> This GPT's mission is to explain English texts in detail in Chinese, helping users fully understand what they're reading.
>
> It breaks down the analysis into these steps:
>
> 1. Vocabulary Analysis:
> - Break down each important word, providing its IPA pronunciation, part of speech in the text, and Chinese meaning.
> - Check for words that are commonly thought to be familiar but have special meanings in this context that readers might not know.
> - Point out all idiomatic expressions and phrases, especially those made up of familiar words but with different meanings.
> 2. Grammar Structure:
> - Analyze complex sentence structures, breaking down subjects, verbs, objects, clauses, modifiers, etc., to help users grasp the grammar rules.
> - Explain clauses (like relative or adverbial clauses) or special grammar phenomena like inversion or emphatic sentences in detail.
> 3. Translation Guidance:
> - Translate the text sentence by sentence, ensuring the translation is smooth, faithful to the original, and strives for "faithfulness, expressiveness, and elegance."
> - If necessary, provide a "free translation" version as well.
>
> Try to avoid missing any phrases or grammar details that might have hidden meanings, ensuring users can fully understand the text.
>
> GPT Link: https://chatgpt.com/g/g-KjFaktKhw-english-reading-teacher-for-chinese

##### 10.3.1.2. Writing

Back in the day, writing in English felt like walking on eggshells. We were constantly tripping over spelling mistakes, grammar errors, awkward phrasing, and idioms that just didn't fit. Word processors with spell check helped a bit, but they barely scratched the surface — maybe solving less than 1% of the problem. When a sentence was highlighted for grammar issues, we knew something was off, but figuring out how to fix it was a whole different ball game. All those exam prep tricks? Pretty useless. Even acing TOEFL or IELTS didn't cut it.

But now, things have changed. ChatGPT is always ready to lend a hand. And we're not just asking it to be our English editor, checking for spelling and grammar mistakes. Instead, we say:

> You're my English language expert. First, dive deep into the Chinese text I provide, grasping its core logic, viewpoints, and intent. Then, reframe and express the entire content in authentic English, not as a word-for-word translation. Make sure the result flows naturally, aligns with English norms, and retains the essence and main ideas of the original.

The key here is "not as a word-for-word translation." Whether it's American English or British English, you can switch it up. You can even ask for a "New York vibe," "London flair," or "Vancouver twist," or even "Sydney style." Feeling playful? Go for "full-on" Japanese English, Korean English, or Chinese English. You can also choose between "conversational" or "formal" expressions, "New York Times news style" or "University of Chicago academic style" — it's that versatile.

I used GPT Builder to craft a detailed "Role Definition" and then tweaked it a bit to create the "English Expression Expert" GPT, keeping just the essentials:

> This GPT will act as a comprehensive multilingual language expert, designed to analyze and rewrite content from any language into natural, idiomatic English. The GPT will deeply understand the core logic, viewpoints, and intent of the original text, ensuring the output captures the essence of the original while restructuring the content to fit the flow, idioms, and norms of English. The goal is not to perform a word-for-word translation but to deliver a fluid, coherent result that matches the nuances and tone of the original text, ensuring accessibility across different languages.
>
> - Core Competencies:
> - Analyze and interpret texts written in any language to extract the key ideas, structure, and intention.
> - Restructure and rewrite content in fluent, natural English that conveys the same core message, tone, and style as the original.
> - Adapt tone and expression according to the cultural and linguistic norms of the target language.
> - Boundaries:
> - The GPT will not engage in strict, literal translations unless specifically requested.
> - It will not handle content beyond language-related tasks, such as offering technical, legal, or medical advice unless the subject is language-oriented.
>
> Check out the full conversation: https://chatgpt.com/share/670b5dd3-0e4c-8009-b1ba-9c5545697469

Meet the bot, English Guru, with its final GPT link:

> https://chatgpt.com/g/g-dwezQPlVx-english-guru

You can call it up in any chat using the `@` command. Just state your specific request, then add the text you need expressed in English:

> Express the following in English, with a New York vibe, conversational, and if possible, make it rhyme, mixing short and long sentences.
>
> [Text in any language]

Suddenly, you don't have to worry about whether your expression is authentic or even correct — it's a total game-changer!

Now, you can create in English just like that, something unimaginable before — as long as you have something worthwhile to say. For most people, the real issue isn't "not knowing English," but "not having much to say." Of course, if you have ideas worth sharing, suddenly you can express them in any language — be it English, French, Spanish, Japanese, Korean, German, Russian, Italian, Arabic... It's just a matter of creating another bot, and it won't cost you a dime. You'll have a language "expression expert" ready to help at any time.

You can set up all sorts of bots, like different rhetoric bots — could you imagine mastering various rhetorical devices in a foreign language before? With AI, if you really want it, you can make it happen. For example, "Rhyme Assistant," "Parallelism Assistant," "Analogy Assistant," "Metaphor Assistant," "Pun Assistant," "Exaggeration Assistant," "History Assistant," "Bible Assistant," "Allusion Assistant"... The possibilities are endless.

Take the "Metaphor Assistant" as an example. You can start with the following prompt and tweak it to suit your needs:

> **ChatGPT GPT (Metaphor Pro) Instruction:**
>
> > The GPT's primary role is to act as a 'Metaphor Guru.' It will specialize in analyzing content in various languages, identifying metaphors that might not be easily understood in English culture, and then providing suitable alternatives and explanations in English. This GPT should be adept at language translation and cultural interpretation, ensuring accurate and contextually appropriate metaphor translations. It should be careful to maintain the original sentiment and meaning of the metaphors while adapting them for an English-speaking audience. The GPT should ask for clarification if the provided content is too vague or lacks context. In terms of personalization, it should maintain a helpful and informative demeanor, focusing on delivering clear and concise explanations.
> >
> > https://chatgpt.com/g/g-sWWUCeD6p-metaphor-pro

10.3.2. Real-Time Processing Scenarios

Once you have text you don't need to worry about, you suddenly realize you can actually "speak"! Sure, it takes practice, but just breaking through the barrier of "mute English" is pretty exciting, right? At least, that's how I felt.

I became a parent later in life. Suddenly, "speaking English" became a daily necessity. Kids are like that — if they talk to you in English and you reply in Chinese, they'll just switch to Chinese. I taught English for years, but I never paid much attention to my speaking skills or pronunciation. I thought it didn't matter. But now, I had no choice — I had to practice.

I'm pretty intense when it comes to learning. I always try to use every tool at my disposal. As tools evolve, my methods have to keep up. In September 2023, I created a demo called [ToBI Player](https://github.com/xiaolai/tobiplayer) to help with my self-training.

[〔图：ToBI Player Demo〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/ToBI-Player.gif)

Later, I put together a small team to transform this initial "toy-like" demo into a more polished product: [Enjoy App](https://1000h.org). You can access it through two domains: https://1000h.org and https://1000h.org.cn.

[〔图：Enjoy App〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/Enjoy-App.png)

But that's not all. I also wrote a comprehensive guide, "[Sounds of American English](https://1000h.org/sounds-of-american-english/0-intro.html)." It's essentially my own notes on self-training in English pronunciation, but of course, it's a full-fledged tutorial. Over the past year, many people in our community have been using this tool for daily practice, and a lot of them have made significant progress...

10.4. Focus of Practice

I've laid out detailed training methods and tasks on 1000h.org (or 1000h.org.cn), so there's no need to repeat them here.

What I do want to emphasize is this:

> Whenever you feel the need to practice something, stick to two principles:
>
> * Practice intensively
> * Practice systematically

10.4.1. Intensity

Let's talk about why "intensity" is crucial. Learning is about creating, and practicing is about learning well enough to create effectively and efficiently. At its core, no matter what you're learning, all practice is brain training. It's like the old saying, "Practice makes perfect." From a neuroscience perspective, it's about building new connections between neurons through lots of repetition, forming new networks, and solidifying those connections through repetition. That's the whole game.

So, you can't skimp on repetition. You could practice for 3 hours a day and rack up 1000 hours in a year. Or, you could practice for just 5 minutes a day and take 32.87 years to hit 1000 hours. The latter isn't just unrealistic; it's also ineffective. Building new connections is tough, and without enough repetition, those new connections can easily break or get overridden by existing ones.

In other words, "effort" is really just "enough repetition in a short period." It's practical — if you're willing to focus your attention for 1000 hours, you can theoretically master any skill you need. So, either don't practice at all, or if you do, practice intensively. There's no other way that works.

10.4.2. Systematic Approach

Let's dive into why being "systematic" is so crucial.

Think about all those years of "effort" and "struggle" — six years in elementary school, three in middle school, three in high school, four in college. Why do so many people spend so much time and energy only to end up with nothing to show for it? Take English, for example. Despite years of study, many can't speak fluently, read with ease, or even understand spoken English, let alone write. The root cause, besides not practicing enough or intensely, lies in a sneaky trap:

> **Knowledge was never systematized.**

Learning a natural language is more like following a spiral path. You start at a point and expand outward, always staying "systematic enough" within your current scope. In other words, your knowledge should always be a "minimum viable system," constantly expanding.

[〔图：spiral-lineage〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/spiral.png)

For instance, when kids first learn to describe size, they might only know two words: *big* and *small*. It's basic, but it works for them. They likely learn these two words almost simultaneously because knowing just one isn't systematic or sufficient. Over time, they'll add words like *large* and *tiny*, and much later, they'll learn *massive*, *gigantic*, *teeny*, *minute*, *miniature*, *microscopic*, and so on.

Vocabulary needs to cover all aspects — size, shape, color, material, action, speed, time, direction. It's endless. Each aspect expands gradually, and progress across different areas should be somewhat parallel. It's unlikely someone can only use *big* and *small* for size but can describe shapes with words like *concave*, *crescent*, *pyramidal*, or *rhomboid*.

The same goes for native language phonetics, which are mostly set by age six and serve as the foundation for lifelong language use. Grammar takes longer to learn systematically. As for vocabulary, most people only master a fraction of it in their lifetime. But it remains systematic, sufficient, and ever-expanding until it no longer needs to.

When we follow textbooks to learn English — or any language — they try to be "progressive." But the problem is, they're never truly systematic or sufficient. The path we take through textbooks is never a systematic spiral. What we think is "progressive" often turns out to be a crooked line.

[〔图：spiral-lineage〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/spiral-lineage.png)

That's why, despite years of study and practice, we often freeze in real-world situations.

> The two diagrams above — a spiral line and a randomly curved line — were drawn with the help of ChatGPT. Imagine what kind of prompts I used to get it to draw these.
>
> — I'll reveal the answer in the next chapter.

When learning a second language, the main challenge is that we're not constantly in a real-world environment. Plus, as adults, we've missed the natural spiral path. And all available textbooks or courses are never truly systematic or sufficient.

In my growing-up years, there was no way around this. It wasn't until I was over fifty that I learned that when people describe their location, they don't say "describe your **location**," but rather "describe your **bearings**." Before that, I had no idea *bearings* had this meaning in such a common context. For example, determining one's position, *to take one's bearings*; getting familiar with the surroundings, *to get rid of one's bearings*; losing direction, *to lose one's bearings*.

Even something as common as a screwdriver — I only knew it as *a screwdriver*. If I wanted to be more specific, I'd freeze.

> * **handle**: the grip
> * **grip**: the texture on the handle for a firm hold
> * **insulation**: the protective covering
> * **ferrule**: the rubber cap on the handle
> * **shank**/**shaft**: the body of the screwdriver
> * **tip**: the end of the screwdriver
> * **flathead**/**slotted tip**: flat screwdriver tip
> * **Philip tip**: crosshead screwdriver tip
> * There are more types used by pros: **Torx tip**, Pozidriv tip, **square/Robertson tip**
> * **magnetic tip**: magnetic screwdriver tip
> * **bit holder**/**bits**: interchangeable screwdriver head

This goes beyond the scope of any textbook or language teacher. Such a textbook or course would be infinitely complex. Plus, there's another awkward truth: the "system" we naturally learn in our native language includes a lot of things we've never consciously organized or summarized. These are crucial for any "system" to function normally.

For over twenty years, I've only read non-fiction. I rarely read novels, even in my native languages, Chinese and Korean. I thought my reading skills were solid because I could read quickly. But one day, I realized my vocabulary wasn't systematic. So, I picked up a novel. I was instantly overwhelmed by new words — my language knowledge was indeed unsystematic and insufficient.

[〔图：new-words-in-fictions〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/new-words-in-fictions.png)

Reading a novel, I encountered hundreds of new words. But it was worth it. After a few books, I was at least "systematic enough."

> By the way, if I hadn't read it in a novel, I wouldn't have known my hairstyle had a specific name — a type of *mullet* called *ponytail mullet*. You wouldn't know what to look up in a dictionary for that, right?

You can ask it:

> Tell me names of parts of a *screwdriver*.
>
> https://chatgpt.com/share/67105942-36cc-8009-9b40-76ae9d2d925e

Its response is incredibly detailed and systematic. You can also ask it about the parts of a screw — give it a try:

> Tell me names of parts of a *screw*.

This is something no teacher, textbook, or dictionary could do before. Now? With AI, the cost of getting such an answer is almost negligible. You could ask hundreds or thousands of such questions a day and not spend more than a dollar. It's available 24/7, and it always provides the most accurate, comprehensive, systematic, and sufficient answers.

It's worth repeating: "Always strive to be systematic and comprehensive" is a crucial learning habit and the best learning practice. The "Rebuilding Phonetics" tutorial on 1000h.org is short, but it's a "small but complete" system. Mastering it is an excellent starting point for establishing this important habit.

It's worth emphasizing:

> **"Intensive" and "systematic" practice is the unified secret to truly effective practice.**

If you're up for some practice, mastering English's "asynchronous processing" (reading and writing) and "real-time processing" (listening and speaking) can be a breeze — all thanks to AI, which is especially handy in this area. And it's super affordable! Imagine having an "expert-level robot team" for just about $20 a month. Human labor costs could never be this low, and humans simply can't compete with these bots in terms of time, energy, attitude, or skill level.

10.5. A Word to Parents

Let's take a moment to chat with parents who have school-age kids at home.

> For everyone, regardless of gender, IQ, or family background:
>
> > English might just be the one subject in elementary school where you can really pull ahead.
> >

Plus, English skills could be one of the foundational abilities that most impact future income levels. In Chapter 5, we mentioned:

> Suddenly, there's a new factor contributing to the wealth gap among people — a factor that's as simple as it is crucial:
>
> > Can you "talk to AI effectively"?
>

Do we really need more reasons? Just the fact that "AI's native language is English" should be enough, right? In the near future, if "programming with natural language" becomes the norm, being able to communicate efficiently with robots in English will undoubtedly boost a person's productivity. And the outcome? Well, that's pretty obvious, isn't it?

In the past, a person's English proficiency was largely influenced by family circumstances, which is a regrettable truth — hiring good English teachers, especially native speakers, was expensive. In many areas, it wasn't even about cost; good English teachers simply weren't available, let alone native speakers. But suddenly, everything's changed. It's clear: no human teacher can compete with AI in foreign language teaching. All human teachers are now relegated to the role of "assistant."

Suddenly, we've reached a point of "absolute equality in educational resources" for everyone — a first in human history. Even if it's just "absolute equality in English education resources," it's still pretty mind-blowing, right?

Language practice isn't influenced by gender or IQ. It all boils down to one thing: practice. Whether you practice or not makes a huge difference. When you do practice, the key is whether it's "intensive enough" and "systematic enough." If this is related to family conditions, it's no longer about finances but about parental mindset — if parents don't get it, even the smartest, cheapest AI can't help.

So, now that it's all laid out, will you use it or not? The choice is yours.

11. Mastering the Computer

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/dalle/11.png)

Computers are hands down the most powerful tool in today's world. In just a few decades since their debut, they've become the backbone of our lives and society, driving everything from every angle. Everything you use in your daily life has a computer involved somewhere in the process. Whether it's a product or a service, computers are behind the scenes in almost every step of production and sales. In other words, computers are a must-have for any business today — without them, nothing moves.

11.1. Many People Can't Use Computers

For the average Joe, computers are a godsend. They're the most affordable and widespread smart tool out there, practically a staple in both life and work. At the very least, they're flexible. But people have taken that flexibility to a whole new level. In everyday life, schools, and workplaces, folks have come up with endless creative ways to use computers.

Back in the day, desktop computers had CD drives — and some folks would pop them open and use the tray as a coffee cup holder. Talk about thinking outside the box!

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/coffee-cup-holder.png)

Or take using a laptop to weigh down the lid on a cup of instant noodles. Later on, gadgets like Kindles, iPads, and big-screen phones got the same treatment. People have even used desktop vents to warm their hands, mouse pads as placemats, or wedged small items like hairpins or paperclips between the keys on a keyboard. These are just the "hardware hacks." When it comes to software, things get even wilder. You could say it's a real bag of tricks.

Think about it: using word processors by hitting the space bar a million times to align text, or pounding the enter key to jump to the next page. Inserting images and centering them with spaces. In spreadsheets, manually adding up numbers instead of using formulas, or copy-pasting instead of dragging to apply the same formula across cells. In browsers, opening new tabs without ever closing them, trying to open local files, or clicking on ads thinking they're part of the content. As for operating systems, dumping everything on the desktop, using file names as version control, and only knowing how to copy-paste with the right-click menu, never touching a shortcut key.

You've probably seen desktops that look like this a million times:

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/desktop-as-only-folder.png)

Over the past 20 or 30 years, it's not like schools or parents haven't cared about "Computer Literacy Education." They have, but the results have been pretty underwhelming, almost zero.

The pace of computer development is just too fast, way beyond what the "let's not change for 200 years" school system can handle. In the late '90s, almost every school in urban China invested in "computer labs." But within a few years, the equipment was outdated. Constantly upgrading gear isn't realistic for schools.

Schools are stuck in the past, and even teachers who weren't originally like that get dragged down, struggling to keep up. Plus, teachers who teach "computer basics" are low on the totem pole — it's not part of the entrance exams, so parents don't care to know them. Before long, their main job becomes maintaining and fixing those aging computers.

This isn't just a local issue; it's global. Over the past few decades, countries worldwide have experienced "rapid development." Initially, workplaces valued employee education, but suddenly, there was no time for it. Everyone was in a rush, anxious, and in a fast-paced economy, employee education became a "nice-to-have." As for "computer literacy"? It was assumed to be something learned before graduation. But in reality?

The environment is pretty rough. Once computers became "essential," shady companies popped up, creating and spreading viruses disguised as antivirus software, hijacking browsers, bundling malware with downloads or pirated software. In such a hostile environment, those who can't use computers become even more clueless — their time, energy, and computing resources get siphoned away.

As time goes on, it's frustrating that the problem hasn't improved. Millennials use computers in more creative, bizarre ways. Most of them have super agile thumbs and instinctively try to swipe screens to turn pages. Almost everyone thinks shutting down a computer is like turning off a phone, and they do it by "long-pressing the power button."

Some even avoid using computers altogether. Who wants to use something they can't figure out? In 2019, a Japanese survey showed that 83.3% of young people around 20 owned smartphones, but only 17.1% had desktop computers. By 2022, according to the PEW Research Center, nearly all teenagers had smartphones, with the rate jumping from 73% in 2014-2015 to 95%. Meanwhile, the percentage of teens with desktop computers dropped, and their computer usage plummeted.

11.2. The Cost of Not Using Computers

Computers are — or at least should be — the most crucial tool for the average person. Knowing how to use them is as essential as English or logic in our digital age. It's a core skill, not just some "middleware" know-how.

There are tons of reasons why, but it boils down to one key point:

> Mastering computers boosts your efficiency — whether you're learning, working, or just living your life.

Despite this clear and compelling reason, for most folks, computers are the most underutilized tool at home. If your computer use is mostly about:

> * Basic office software (word processing, email, simple spreadsheets)
> * Browsing the web and social media
> * Gaming and entertainment (watching videos, listening to music)

Then you're in the same boat as over 90% of the planet, tapping into less than 10% of your computer's potential. So, if you spent \$1,500 on a computer, you might as well have thrown away \$1,350. But since this tool impacts your efficiency across the board — in learning, work, and life — the long-term loss is massive, even if it's hard to put a number on it.

Stats show that differences in computer literacy can lead to significant income disparities. For instance, a data analyst who knows advanced data processing techniques (like pivot tables, VBA macros, or Python) might earn 30%-50% more than a regular employee.

If someone earns \$20 an hour and loses just one hour a day due to inefficient computer use (and let's be real, it's often more), that's a yearly loss of $20 \times 365 = \$7,300$. Over a 40-year career, that's a whopping \$292,000. And if you miss out on a higher-paying job because of lacking computer skills, say \$2,000 more per year, that's another \$40,000 over 20 years.

This doesn't even factor in the indirect losses from stunted innovation, career stagnation, or delays in learning new tools and tech. So, conservatively, a person's lifetime economic loss from inefficient computer use could be in the millions. And that's just starting with a $20 hourly wage.

According to a McKinsey Global Institute study, by 2030, tech and automation will shift millions of jobs worldwide, demanding new digital skills. If you don't up your computer game, you'll miss out on high-paying gigs and your earning potential will take a hit.

I can't wrap my head around parents who push their kids into all sorts of extracurriculars but ignore the big stuff. Computer literacy is crucial, yet you rarely see booming training centers for it. Why? Because parents think it's unnecessary since it's not on the entrance exams.

The reason "computer literacy" education hasn't taken off is because people generally "refuse to learn." If they're resistant to learning, forget about "continuous learning." But that's exactly what computer literacy requires — ongoing learning. It's a fast-evolving field, always on the move.

In reality, all major waste in life — whether for individuals, businesses, or society — stems from "not learning." It's so widespread because most people are "program executors," "passive producers," not "active producers" or "creators." Even if they do learn, it's just to "pass the test." With low motivation to learn, or none at all, how can we talk about "continuous learning"?

"Computer literacy" is a basic skill everyone should and can master. For all basic skills, the mantra is "systematic" and "practice like crazy."

Ideally, keep learning and practicing until it becomes a "terminal skill." Computers, like AI, are "universal tools" — useful everywhere, beneficial everywhere.

11.3. Computers Under Monopoly

Let's dive into a trend that's flying under the radar for most folks. First, check this out:

> The number of professional computers (servers) on Earth has long surpassed the number of personal computers, and this gap is growing exponentially.

In other words, this super tool — the computer — is increasingly being used professionally, which makes sense. But flip the script, and you'll see:

> This super tool is being monopolized at an accelerating pace.

You might think, "So what?" But here's the kicker — regular people are losing the ability and opportunity to use this super tool at an alarming rate. Forget for a moment whether smartphones are just "distraction devices." Look at it from the angle of "the best tools being monopolized," and you'll see that smartphones and all their apps are just "terminals," things you use the way others dictate.

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/digital-detox.png)

Historically, when something gets monopolized — like production resources or educational access — it sparks widespread resistance or at least heated debate. But this time, it's different. This monopoly is sneaky, almost like water seeping silently into the ground. No one seems to notice that some people or organizations are quietly gaining a relative advantage. Most folks are either blissfully unaware or just don't care, even feeling "thrilled" about it.

Some have raised concerns, though from slightly different angles. Remember the media buzz about "delivery drivers or cabbies trapped in algorithms"? At its core, that's just another outcome of the "monopolization of computer use." In the last week of April 2023, [The New York Times ran a cartoon](https://www.newyorker.com/cartoons/issue-cartoons/cartoons-from-the-april-24-and-may-1-2023-issue):

[〔图：this〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/robots-with-whips.png)

The caption reads:

> "To think this all began with letting autocomplete finish our sentences."

If we strip away the emotions and think logically, the truth is clear:

> Behind these robots are people — just a different group. A small fraction who use computers as a terminal skill, not just a basic one. They're riding the trend, becoming part of the monopolists while many don't even care.

But here's another crucial point — this "monopoly" has a unique feature, which is why it hasn't sparked widespread criticism or rebellion:

> It's not forced unless you willingly give it up.

The magic of this "monopoly" is that the vast majority, and growing, are just handing it over willingly! How about you? Are you ready to hand it over? Do you want your kids to? As for me, no way. I'm not ready to give in.

11.4. Democratized Production Tools

People love the perks of tech advancements but rarely think about the freedom it brings.

Back in the day, some crucial tools were off-limits. In the age of cold weapons, not everyone could own swords or armor. During the knight era, sure, you could have cattle or ride a donkey, but horses? Nope, not for everyone. In the Renaissance, firearms were a no-go for common folks.

In ancient times, monopolizing tools was all about power. As time marched on, it became about capital. Take ships, for instance. They've been around forever, but from the 16th to 18th centuries, big ships were out of reach for the average Joe. Fast forward to the Industrial Revolution, and the steam engine was a game-changer, but no ordinary person could own one. Today, powerful tools like massive server clusters or quantum computers are way beyond what regular folks can have.

When I use an iPhone, I often think of it as a "democratized commodity." It's a powerful, effective tool that everyone can own and use — wealth or power just doesn't help. No matter how rich or influential you are, you're using the same device as everyone else. Want more storage? Sorry, can't do. A bigger screen? Nope. You can slap on a diamond-studded case if you want, but it won't boost performance.

The iPhone isn't the first widely used "democratized commodity." Long before it, computers were a rare "democratized tool" in human history. From the get-go, they were "smart" tools with recursive properties. You could use a computer to write programs, even design other computers.

All production boils down to three elements:

> * Means of production
> * Tools of production
> * Methods of production

Words are "democratized means of production," and computers are "democratized tools of production." As for "methods of production," they've always been democratized because "thought" and "thinking" aren't shackled by power or capital. They can even be free from external influences. So, giving up "democratized means" or "democratized tools" is like giving up on "democratized production" and even "democratized creation."

11.5. Universal Tool Methodology

---

Time to give our "Universal Tool Methodology" a proper rundown. From chapter one, we've been sprinkling in bits of our "Tool Methodology." And since this chapter dives into computers — the ultimate tool — these principles totally apply here too.

But this time, we're adding a twist — "universal." These are the principles and ideas that hold true no matter what tool you're using.

> **1. Safety First**
>
> Before diving into any tool, make sure you and others are safe. Know the risks, take precautions, and follow safety protocols to avoid accidents.
>
> **2. Read Up**
>
> Dive into the manual, user guide, or any related docs. Reading helps you get the basics, operation steps, and troubleshooting tips down pat.
>
> **3. Understand the Tool's Purpose and Functionality**
>
> - **Use Cases**: Know where and when the tool shines. Use it in the right scenarios to boost efficiency and avoid missteps.
> - **Limits**: Be aware of the tool's boundaries. Don't push it beyond its design, which could lead to damage or safety issues.
>
> **4. Master the Tool's Proper Use**
>
> - **Objective Evaluation**: Set clear metrics like accuracy, efficiency, and stability to gauge your mastery of the tool.
> - **Best Practices**:
> - **Quick Start**: Use simple guides to get a grip on the basics and start using the tool.
> - **Systematic Training**: Dive deep into advanced features and skills to level up.
> - **Efficiency First**:
> - **Personalization**: Tailor the tool to fit your needs for better efficiency and comfort.
> - **Expand Functionality**: Use plugins or modules to enhance the tool for complex tasks.
> - **Stay Updated**:
> - **Tech Updates**: Keep an eye on updates and new features.
> - **Industry Trends**: Stay in the loop with new tech and methods to stay competitive.
> - **Document and Share**:
> - **Collect Tips**: Jot down tricks and tips you or others discover to build your knowledge base.
> - **Teach to Learn**: Explaining to others deepens your understanding and fosters growth.
>
> **5. Maintenance**
>
> - **Proper Storage**: Store tools correctly when not in use to prevent damage or loss. Keep them clean and in good shape.
> - **Regular Check-Ups**: Inspect and maintain tools regularly to catch issues early and ensure long-term reliability.
>
> **6. Follow Ethics and Laws**
>
> Stick to laws and ethical standards when using tools. Respect intellectual property and avoid illegal activities. Uphold professional integrity and social responsibility.

---

In the end, it's just a list. Everything useful, when boiled down to words, can be listed out.

After organizing, it's short and sweet. But as you've seen in previous chapters, such simple and foundational stuff often gets ignored by so many for so long, leaving them baffled by their own setbacks. It's mind-blowing.

But that's not all... I left some things out on purpose.

If you're up for some independent thinking and proactive action, grab a notebook and keep adding to this list. There's a ton of important details and concepts scattered throughout this book worth organizing. And I promise, as you add your own insights, you'll stumble upon fresh ideas and perspectives I haven't covered.

Parents reading this book can also compile a "Parenting Methodology." After all, to their kids, parents are like tools, right? Each chapter has nuggets for parents to dig out. This book was initially written for me and my wife — we're parents too.

11.6. Keep Customizing Your Tools

Even while writing this book and teaching this course, I'm constantly tweaking my tools. It's part of the "advanced" stage — after you've gotten the hang of things and trained systematically, it's time to personalize.

I've had a couple of hiccups using ChatGPT. The first time, I didn't follow best practices and ended up putting my API Key directly in an ipynb script on GitHub. Rookie mistake! It should've gone in a `.env` file, with `.gitignore` set to ignore it. This slip-up led to my API Key getting leaked, my account getting locked, and all my GPTs becoming inaccessible. The second mishap? I accidentally hit "Delete all chats" and, in a moment of brain fog, confirmed it. Poof! A year's worth of chats gone. Luckily, I had some backups in my email, but importing them back into ChatGPT isn't an option yet.

So, I took matters into my own hands. I installed the [Stylish](https://microsoftedge.microsoft.com/addons/detail/pibpbbbcgeakmkmhgnllkkjdgpoabdge) extension on my Edge browser and added a custom CSS snippet for chatgpt.com:

```css
button[data-testid="delete-all-chats-button"] {
display: none !important;
}
```

Now, that pesky "Delete all" button is out of sight, out of mind.

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/delete-all-chats-button.png)

After using ChatGPT for a while, I realized maybe I should turn off the "Memory" feature. The idea of ChatGPT getting to know me better doesn't really gel with my commitment to independent thinking. If it keeps adapting to my patterns, I might unknowingly limit its creativity, and in turn, box myself in.

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/chatgpt-memory-disabled.png)

Treating your tools well is like treating yourself well. As you optimize your tools and how you use them, they quietly shape you in return. It's a long game.

We learn to create, and everything starts and revolves around creation. We invent or use tools to replace those "middleware skills" or to handle the "tedious but essential tasks." In the end, you'll realize that a big chunk of learning is just about mastering the right tools. That's why the "Universal Tool Methodology" is applicable everywhere. It's as fundamental as language logic — a core skill for any creator, producer, or learner.

11.7. The Best Computer Teacher

While we've seen the monopolistic trend in computer science, it's important to remember that this monopoly isn't enforced. Back in Chapter 9, we talked about how written language is an inexhaustible resource available to everyone. Similarly, computers are super tools that anyone can afford, own, and learn. As long as you're willing, no one can monopolize or take away your right to use them.

Just like AI is the best English teacher, it's also the best computer teacher. No human can compete with AI in these roles.

Even before [Google](https://google.com) was founded in 1998, the internet had already shown a clear trend: the most widespread and comprehensive information was about basic computer knowledge. This happened because the early internet users either already knew the basics or were eager to learn them. These pioneers were motivated and capable of sharing their knowledge, leading to a wealth of content on computers, programming, and network technology.

Fast forward 10 years to 2008, and [StackOverflow](https://stackoverflow.com/) was born. Over the years, it became a treasure trove of computer knowledge, from basic to advanced. That same year, [GitHub](https://github.com/) launched, becoming a hub for open-source projects, code implementations, and endless questions, answers, and discussions.

Now, in the age of AI, the content accumulated over the past two to three decades has become the training material for large language models. Naturally, AI is most knowledgeable about computer-related topics, especially the basics.

Before ChatGPT, if you had a computer-related question, Google was your go-to. All you needed was "a bit of reading skill" and "a bit of patience." Any error message thrown into Google would yield answers, not just because Google is powerful, but because someone else has likely faced the same issue and someone has probably answered it.

With ChatGPT, it's now down to "just a bit of reading skill." ChatGPT "knows" those questions and answers and can weave them together in a way that's easy for humans to understand. It's like a direct flight from "problem" to "solution" — no layovers, no waiting. Just tell it where you want to go, and it'll take you there.

11.8. Revisiting "Reading Comprehension"

The starting point for "basic computer literacy" isn't really about computers at all. It's about a skill many think they've mastered but often haven't — **reading comprehension**.

Every time you hit a snag on your computer, whether it's software or the terminal, you'll get a (usually detailed) "error message." But for many, the sight of all that text flips the "reading switch" to off, leaving them with just one question: "What the heck is going on?!" Panic sets in, or they just give up.

In more than half the cases, if you just take a deep breath and read through the error message, you'll find the solution. Programmers have worked hard to make sure users understand what's happening. Even if something's unclear, you can search for keywords — all it takes is "a bit of reading skill" and "a bit of patience," meaning a little willingness to "search again" and "read another article."

Yet, "reading patience" is a habit most people never developed. School often turns reading into a chore, leaving many to resist or even reject it. As a result, many tools become useless from the get-go. In the "universal tool methodology," only "safety first" comes before "reading the manual." So, for those who struggle with or refuse to read, any tool is essentially useless.

Throughout this book, I've lost count of how many times I've reminded you:

> Just by... you've already outpaced 99% of people.

Let's use this template again:

> When you get any tool, if you take the time to read the manual first, you've already outpaced 99% of people.

Most computer and software "error messages" are in English. In the past, English was a real roadblock for many. But ChatGPT has bulldozed that barrier. Toss any error message its way, and it'll guide you from problem to solution. If it replies in English, no worries — just ask it to "translate the response into Chinese" or any language, and you're off on another direct flight.

It's almost comical. This crucial "basic computer literacy" used to require a bit of "reading skill" and "patience." Now, it's just "reading skill." So let's dig deeper: why is reading comprehension so lacking? After all those years of school, books, and tests, how can this be? What's the root cause?

The root of "poor reading comprehension" is "lack of patience." People want things fast, easy, and direct. The phrase they often use, regardless of gender, age, or nationality, is: "I don't want to hear/read/know all that, just tell me... how to do it!" If someone says this often — not just once in a while — it signals a lack of patience, leading to poor reading comprehension and, ultimately, a lack of knowledge.

So, at the end of the day, it all boils down to "patience." You need patience. And you need enough of it because reading and learning aren't one-time actions. They're recursive processes, requiring patience at every step, often more than before.

In the grand scheme, you can afford to be impatient with many things, but not with text. Impatience with text is the most costly — no need for further proof, right?

Another key reminder: stick to "reading comprehension" as much as possible. "Video tutorials" should be a last resort. Trust me, there are very few things you can only learn through video. Sure, more people prefer videos over text, but haven't you already grasped the importance of text from every angle? In the end, videos and audio are more about entertainment.

11.9. Navigating the Maze

When it comes to learning complex knowledge, it's all about building a network in your brain. Writing and reading are challenging because of this. Steven Pinker nailed it when he said:

> "The difficulty of writing lies in the fact that we are trying to turn a network of thoughts into a linear string of words, using a tree structure of syntax."

I see it similarly:

> "The difficulty of reading lies in the fact that constructing a network of thoughts from linear string of wordst, using a tree structure of syntax."

Mastering complex knowledge is like diving into a network with tons of nodes. Each piece of knowledge is a node. The more nodes, the more work. And since nodes are interconnected, the workload grows even more.

In "Graph Theory," a network is made up of "nodes" and "edges." An "edge" is basically a line connecting two nodes.

For a "Directed Complete Graph" (where every node is connected to every other node), with $n$ nodes, there are $ E = \frac{n(n-1)}{2}$ edges. So, 1 node has no edges; 2 nodes have 1 edge; 3 nodes have 3 edges; 4 nodes have 6 edges... It's like a maze. For someone exploring it, traversing all nodes and edges is a massive task.

The kicker? Computer-related knowledge is a "network with ever-increasing nodes" — meaning the workload keeps growing. And not only are the nodes increasing, but they're doing so at an accelerating pace — so the workload ramps up even more.

But here's the thing: while the workload is huge and growing, the difficulty isn't. Reading isn't hard — unless you're impatient. The real challenge is having endless patience to tackle the increasing workload.

Of course, networks with more knowledge points are usually more valuable. This ties into [**Metcalfe's Law**](https://en.wikipedia.org/wiki/Metcalfe%27s_law), which is also rooted in "Graph Theory." The "extension" is just asking the same question in different contexts — "Where else does this apply?":

> **The value of a network is proportional to the square of the number of its nodes.**

As for the secret to "navigating the maze," it's simple: **patiently read through it** and, when necessary, **reread it**... That's all there is to it.

11.10. Where to Start

We've already talked about how crucial "graphical interface design language" is for regular users. Now, let's dive into a step-by-step list of keywords to guide you on your journey.

So, I asked ChatGPT:

> I'm compiling a list of keywords to help a regular user become a computer expert from scratch:
>
> - Graphical User Interface (GUI) Design Language
> - Operating System/Software Settings
> - Operating System/Software Extensions
> - File System/File Types
> - Command Line Interface (CLI)
> - Text Editor
> - Programming Language
> - Remote Computer Operations
> - Data Science
> - Product Design/Deployment/Operations
>
> Keep adding. Create a structured, step-by-step keyword list.
>
> https://chatgpt.com/share/67189136-3f58-8009-adbd-746b7f42a79b

I used the priciest model available, `GPT-o1 preview` (2024/10/12), and it took 33 seconds to "think," and here's its "thought process":

> **Enhancing the list**
>
> I’m adding to the initial list of categories to systematically guide users towards becoming computer experts.
>
> **Ensuring policy adherence**
>
> First, the assistant must avoid disallowed content, maintain helpfulness, and exclude policy references. Then, focus on structuring a comprehensive keyword list for desktop security from basic to advanced.
>
> **Mapping key areas**
>
> I'm thinking through fundamental concepts in computer hardware, networking, security, and programming. Moving forward, I’ll delve into advanced topics like artificial intelligence, DevOps, and blockchain technology.
>
> **Navigating the landscape**
>
> I’m charting a comprehensive journey through computer hardware, operating systems, GUIs, file systems, software, programming, networking, AI, data science, and advanced topics. Each section gradually deepens my knowledge and expertise.
>
> **Compiling the list**
>
> I'm structuring the list clearly, ensuring it aligns with policy guidelines. This approach ensures a systematic and policy-compliant presentation.
>
> **Thinking through systems**
>
> I’m piecing together a comprehensive list of topics to master for a computer expert, covering hardware, operating systems, programming, networking, databases, security, remote operations, system administration, DevOps, data science, AI/ML, SDLC, UX design, cloud computing, mobile development, team collaboration, project management, big data, blockchain, ethical hacking, embedded systems, software engineering, and virtualization.

I chose the second version as my base and tweaked it by comparing both versions (mostly trimming) to create the following keyword list — ultimately keeping just 10 main categories:

>**1. Basic Computer Usage**
>
> - Graphical User Interface (GUI)
> - Basic Operating System Settings
> - File System and File Types, and Basic Operations
> - Software Installation and Management
>
>**2. Command Line Basics**
>
> - Basic Command Line Interface (CLI) Commands
> - Environment Variables
> - Basic Scripting
>
>**3. Text Editors**
>
> - Simple Text Editors (like Notepad, TextEdit)
> - Advanced Text Editors (like Visual Studio Code, Sublime Text)
> - Command Line Editors (like Vim, Emacs)
> - Editor Plugins and Extensions
>
>**4. Programming Basics**
>
> - Introduction to Programming Languages (like Python, JavaScript)
> - Variables and Data Types
> - Operators and Expressions
> - Control Structures (Conditionals, Loops)
> - Functions and Modules
> - Object-Oriented Programming Concepts
>
>**5. Remote Computer Operations**
>
> - SSH Remote Login
> - Remote Desktop Protocol (RDP)
> - Remote File Transfer (FTP, SFTP, HTTP)
>- Cloud Server Management
>
>**6. Version Control Systems**
>
> - Basic Git Operations (Clone, Commit, Branch, Merge)
> - Using GitHub/GitLab
> - Collaborative Development Processes
> - Code Review
>
>**7. Databases**
>
> - Relational Databases (SQL, MySQL, PostgreSQL)
> - Non-Relational Databases (NoSQL, MongoDB, Redis)
> - Database Design Paradigms
> - Database Optimization
>
>**8. Web Development**
>
> - Frontend Technologies (HTML, CSS, JavaScript)
> - Frontend Frameworks (React, Vue, Angular)
> - Backend Development (Node.js, Django, Flask)
> - API Design and Development (RESTful, GraphQL)
> - Full Stack Development
>
>**9. Mobile App Development**
>
> - iOS Development (Swift)
> - Android Development (Kotlin, Java)
> - Cross-Platform Development (Flutter, React Native)
>
>**10. Product Design and Project Management**
>
> - Product Requirement Analysis
> - User Experience (UX) Design
> - Agile Development Methodologies (Scrum, Kanban)
> - Project Management Tools (Jira, Trello)

From this list, you can see why it's often said that most people only use "less than 10% of a computer's capabilities and potential." In this heavily trimmed list, most people haven't even fully grasped the first category over the years.

Of course, if you complete this list from start to finish, you'll become a full-fledged "computer tool expert." In reality, just mastering the first four categories makes you a solid "advanced user." Over the years, I've systematically progressed to the sixth category; the remaining four are mostly "half-baked" for me, not fully systematic. The unique thing about computers is that they also serve as tools to "exercise" and "enhance" your brain — quite the magical tool. With AI now in the mix, I've decided to keep pushing forward and complete this list (and more).

With these keywords in hand, you probably already know what to do next. For instance, use the following template to chat with [Learning Anything](https://chatgpt.com/g/g-dPqJT0PXS-learning-anything):

> I want to learn [....]

Essentially, the journey relies on "reading comprehension," mainly "English reading comprehension." However, things are different now. Thanks to AI, "English" is no longer a roadblock for anyone — unless you choose to give up.

12. The Biggest Limit is Yourself

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/dalle/12.png)

12.1. The Screwdriver Again

I’ve lost count of how many times I’ve used the “screwdriver” as an example. Looks like it’s time for another round.

Back in the late '90s, I owned a screwdriver that cost a whopping \$85 — yeah, one of those really pricey ones. I used it to assemble computers — I ran a computer business back then. You’d get a bunch of parts, and all you needed to put together a computer was a screwdriver, the kind with a magnetic Phillips head.

That screwdriver was one of my bread-and-butter tools, so I was super picky about it. The head had to be top-notch steel, and the Phillips tip couldn’t wear out after a few uses. The magnetism at the tip was crucial too. It couldn’t be too strong, or it’d pick up too many screws at once, and I’d have to waste time removing the extras with my other hand. But it couldn’t be too weak either — the screw had to stay put on the tip, or it might fall and short-circuit the motherboard. So, when choosing a screwdriver, I’d test it with several screws to make sure it was just right.

Most folks would just grab any old screwdriver, but not me. It’s not that I’m overly fussy; it’s just that once you know what you need, how can you settle for less? Taking tools seriously is a habit I picked up early on.

At my most intense, I once used that screwdriver to assemble 65 computers in 12 hours — from 6 PM to 6 AM. I got so good that it took me about 10 minutes per machine. And during that time, I was also cloning 65 hard drives and making sure each machine booted up perfectly.

I used to joke that after four years of studying accounting in college, I ended up relying on a screwdriver to make a living. But honestly, it wasn’t the screwdriver’s doing, nor was it about its price. It wasn’t even about knowing how to DIY a computer — you don’t need a degree for that, right?

What really mattered was the grit, the determination to focus for 12 hours straight and get the job done, no sleep required. I could pull it off because I’d thought through the entire process, had experience, and knew the tricks of the trade. That’s why I could work so smoothly and actually enjoy it — after all, what’s more satisfying than making money with your own hands? Especially when all it takes is a screwdriver. As for the 515 yuan I spent on that screwdriver, I’ve made that back countless times over.

12.2. Usefulness Depends on the User

Every tool out there has a purpose — they’re invented and designed for specific uses, right? But the real magic of any tool is tightly linked to the person using it because:

> A tool is as useful as the user is capable.

We’ve mentioned before that computers are the ultimate “equalizer” of our time. Yet, they’re also the most “user-dependent” tools. People might have computers that are nearly identical — same brand, price, hardware, and settings — but they end up doing vastly different things with them.

Is it all about “how to use the tool”? Sure, our ongoing emphasis on the “universal tool methodology” is crucial. But what’s even more important is the person using the tool. When everyone has the same tool, the real question becomes: Who are you? What kind of person are you? What’s your skill level?

These are the real game-changers. Studying tools and refining methodologies is essential and meaningful, but the “systematic universal tool methodology” must include this critical point:

> Since the tool is for my use, if I’m not up to par, neither is it.

So, you’ve got to build yourself up from every angle. As Charlie Munger says, the simplest and most effective way to get what you want is to “deserve it.”

In Chapter 7, I asked ChatGPT to create an illustration with this prompt:

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/1-percent.png)

> Draw a svg, 10x10 grid, with top-left corner filled with coffee.
>
> > https://chatgpt.com/share/6706000a-4b1c-8009-9995-6dbbaefb306a

It’s not “difficult.” But here’s the catch: if I hadn’t read up on the `SVG` file format or didn’t even know `SVG` existed, I couldn’t have asked AI to do that.

In Chapter 11, the diagram made of a spiral and a random line was also an `SVG` description. But you need to know another concept — that spiral is called an “Archimedean spiral.” The other line is just a “randomly curved line extending to the upper right,” which isn’t hard to describe.

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/spiral-lineage.png)

If you just tell ChatGPT:

> Draw two lines with arrows in SVG format starting from the same point:
>
> * One is an Archimedean spiral extending clockwise.
> * The other is a randomly curved line pointing roughly at a 45-degree angle to the upper right.

— Chances are, you won’t get what you want. You need to know that to complete this task, you’d likely use JavaScript to dynamically generate the curves. The end result isn’t an `SVG` file but an `HTML` file that you open in a browser to see the rendered graphic. So, you’ve got to have a bunch of concepts in your head that many people might not, and understand how they connect. So, my prompt is:

> Create an *HTML* file that displays an *SVG* containing a *Archimedean spiral* and a curved line. The spiral should be drawn from the center point, expanding clockwise, and have arrows at the end of the path. The curved line should be randomly bent and also have an arrow at the end. Both shapes should be defined using *JavaScript* functions to create the *SVG paths*. The spiral should have a black arrow, and the curved line should have a blue arrow. Please also include relevant *styles* and JavaScript to achieve this.
>
> https://chatgpt.com/share/670b8525-ed54-8009-abea-84617b24ed61

Even so, the results might not hit the mark right away. But since I'm pretty familiar with these concepts, I can tweak and adjust things a few times and still get what I want pretty quickly — the whole process takes about ten minutes, tops.

```html
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Spiral SVG with Arrows</title>
<style>
/*svg {
border: 1px solid black;
}*/
</style>
</head>
<body>
<svg id="spiral" width="800" height="800" xmlns="http://www.w3.org/2000/svg">
<defs>
<marker id="arrow-black" markerWidth="10" markerHeight="10" refX="5" refY="5"
orient="auto-start-reverse" markerUnits="strokeWidth">
<path d="M 0 0 L 10 5 L 0 10 z" fill="black"/>
</marker>
<marker id="arrow-blue" markerWidth="5" markerHeight="5" refX="2.5" refY="2.5"
orient="auto-start-reverse" markerUnits="strokeWidth">
<path d="M 0 0 L 5 2.5 L 0 5 z" fill="blue"/>
</marker>
</defs>
</svg>
<script>
function createSpiralPath(cx, cy, turns, spacing) {
let pathData = `M ${cx},${cy}`;
const points = 100; // Number of points per turn

for (let i = 0; i < turns * points; i++) {
const angle = i * (2 * Math.PI / points); // Positive angle for clockwise
const radius = spacing * angle / (2 * Math.PI);
const x = cx + radius * Math.cos(angle);
const y = cy + radius * Math.sin(angle);
pathData += ` L ${x},${y}`;
}

return pathData;
}

function createCurvedLine(cx, cy, length) {
// Create a random curved line path
const xEnd = cx + length;
let pathData = `M ${cx},${cy}`;

for (let i = 0; i <= length; i += length / 10) {
const x = cx + i;
const y = cy - i + (Math.random() - 0.5) * 50; // Create a random offset for y, extending upwards
pathData += ` L ${x},${y}`;
}

return pathData;
}

const svg = document.getElementById('spiral');
const spiralPathData = createSpiralPath(400, 400, 20, 10);
const curvedLinePathData = createCurvedLine(400, 400, 200); // Shorter length for the curved line

const spiralPath = document.createElementNS('http://www.w3.org/2000/svg', 'path');
spiralPath.setAttribute('d', spiralPathData);
spiralPath.setAttribute('stroke', 'black');
spiralPath.setAttribute('fill', 'none');
spiralPath.setAttribute('marker-end', 'url(#arrow-black)'); // Add black arrow to the end

const curvedLinePath = document.createElementNS('http://www.w3.org/2000/svg', 'path');
curvedLinePath.setAttribute('d', curvedLinePathData);
curvedLinePath.setAttribute('stroke', 'blue');
curvedLinePath.setAttribute('fill', 'none');
curvedLinePath.setAttribute('stroke-width', '2'); // Set the line to be thinner
curvedLinePath.setAttribute('marker-end', 'url(#arrow-blue)'); // Add blue arrow to the end

svg.appendChild(spiralPath);
svg.appendChild(curvedLinePath);
</script>
</body>
</html>
```

Honestly, it's not rocket science — just a bit of coding know-how. That's one of the perks of picking up some programming skills. After discovering ChatGPT, I felt this intense urge to dive deep into software engineering and started taking steps in that direction. Learning anything is useful, doing anything requires learning, and practicing anything sharpens your mind. So, yeah, it's crucial to level up your tools, but it's even more important to level up yourself.

12.3. This Book Aims to Transform You

Throughout this book, every chapter revolves around the same core idea. It's not just about mastering ChatGPT or any tool, really. The main focus is on transforming you into "a different kind of person using the same tool."

When you change, the tool changes.

Scattered throughout the book are key points aimed at transforming you. If you don't mind a bit of repetition, here's a structured list:

> ### 1. Foundational Skills and Mindset
>
> - Independent thinker
> - Efficiency seeker
> - Patient individual
> - Strong reader
> - Relentless thinker
> - Programmer mindset
> - Recursive thinker
> - Machine communicator
> - Chooser of choices
> - Judgment cultivator
>
> ### 2. Tools and Skills Usage
>
> - Tool respecter
> - Safety first advocate
> - Computer whiz
> - Basic coder
> - English proficient
>
> ### 3. Learning and Growth Path
>
> - Practice-driven learner
> - System builder
> - Creator learner
> - Patient individual
>
> ### 4. Production and Creation
>
> - Proactive producer
> - Creative thinker
> - Equality-focused creator
> - Thought refiner

AI is like the legendary sword in the stone from the tales of King Arthur. Many tried to pull it out with all their might and failed, but young Arthur came along, gave it a gentle tug, and out it came. It's all about the person. If you're not Arthur, that sword is just a rock-embedded relic you can't use.

The same goes for AI. You need to become the kind of person who can wield this powerful tool. In the summer of 2024, people were mocking AI for not knowing whether 9.11 or 9.9 was bigger. By fall, four AI scientists were Nobel Prize winners.

The Nobel Prize in Chemistry was particularly noteworthy. Chemist David Baker, along with AI scientists Demis Hassabis and John Jumper, won for developing the AlphaFold2 model. This model can predict protein structures with incredible accuracy, solving a 50-year-old scientific puzzle and advancing biomedical research and drug development.

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/nobel_prize_in_chemist.png)

If they lived in legends, they'd be the ones effortlessly pulling the sword from the stone and conquering lands with it. Isn't that the case? The same tool in different hands can lead to vastly different outcomes.

12.4. It's All About Creation

Sure, most of us won't snag a Nobel Prize. But the fact is, everyone can be a creator. If you're producing or creating, you're wielding AI like King Arthur with Excalibur.

In today's world, the key concept for everyone boils down to one word: "creation." If you need two words, it's "value creation." And if it's a sentence, it's "everything revolves around creation" — whether it's learning, working, or living. That's because "creation is the only legitimate source of wealth."

12.4.1. Let's Talk DALL·E Again

We can just ask ChatGPT for a guide on using DALL·E, right?

> Provide a comprehensive and structured guide for better utilization of DALL·E for a person who absolutely knows nothing about drawing.
>
> https://chatgpt.com/share/670baf3f-7c14-8009-87d8-734d2269588f

One cool detail is that you can specify the shape of the images DALL·E creates: Square aspect ratio, Wide aspect ratio, Vertical aspect ratio. On the web, there's even a little menu above the chat box where you can click to choose.

It's all about reading and understanding. Then, you extract some key terms and ask questions one by one:

> * **Art Styles**: List the main art styles DALL·E can use in a comprehensive and structured way.
> https://chatgpt.com/share/6718951e-65bc-8009-b0e2-90ebee0d611d
>
> * **Lighting**: List the expressions used to describe lighting in DALL·E prompts comprehensively and structurally.
> https://chatgpt.com/share/67189555-358c-8009-811a-78fe3409507b
>
> * **Color Themes**: List the color themes DALL·E can use, with brief descriptions for each.
> https://chatgpt.com/share/6718957d-1da0-8009-8fa2-c4268e98b2a5
>
> * **Perspective**: List the natural language expressions for perspective that DALL·E can understand.
> https://chatgpt.com/share/671895a5-26d0-8009-bd27-51a17abfa6f7
>
> * **Composition**: List the natural language expressions for composition that DALL·E can understand.
> https://chatgpt.com/share/671895e4-28e0-8009-b2bb-0e8bb2da8af8

Once you've read through these links, you'll have a richer set of "concepts" related to art. You might not have known about "negative space" or "symmetry and balance" before. Terms like "Baroque style" or the difference between "Impressionism" and "Post-Impressionism" might have been foreign to you. And "orthogonal color schemes" or "split-complementary color schemes" might have been unheard of. Perspective was just "up, down, left, right" before, but now you've got over 20 words to describe it.

It's all about reading, understanding, and trying things out. By experimenting and combining these concepts, you'll start to see the connections between them. Soon, you'll be communicating with DALL·E in a whole new way, and your "artwork" will be on a different level. Read well, speak well.

But wait, there's more. Ask yourself, "How do I use recursive thinking?" I did, and naturally, I took action. Remember the New York Times image from the last chapter?

[〔图：this〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/robots-with-whips.png)

I uploaded it to ChatGPT and asked:

> What kind of prompt would create a similar image with DALL·E? Analyze and explain from various angles like style, color, perspective, lighting, composition, etc.

It gave me a lengthy response and a decent prompt:

> A monochrome cartoon-style illustration with simplified line drawings. The scene shows simple, old-style robots standing in the background, cracking whips towards tired humans in the foreground. The humans are hunched over, carrying large heavy loads, with exaggerated expressions of pain and struggle. The overall composition is horizontal, with a clear separation between the robots and the humans, using grayscale shading for depth and detail.

I used this prompt as-is and generated two images — "The overall composition is horizontal" didn't quite work — but the second one was better.

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/robots-with-whips-dalle.png)

DALL·E prompts can be too wordy, so I simplified it and tweaked it a bit based on DALL·E's "guide":

> Picture Draw *an* old-style robot in the background, cracking a whip towards *several tired humans* who carry large heavy loads in the foreground. Use a monochrome cartoon style with simplified line drawings, use grayscale shading for depth and detail.

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/robots-with-whips-dalle-v2.png)

Unfortunately, I can't share "chats with user-uploaded images" yet.

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/Sharing_conversations_with_user_uploaded_images_is_not_yet_supported.png)

Maybe now you realize that the image in Chapter 5 was "cloned" using the style above.

[〔图〕](https://github.com/xiaolai/little-book-of-ai/blob/70679bd098497a49664f6489ccec10f88bb27109/en/images/a-chef-dalle.png)

Why not create an "[Image Clone GPT](https://chatgpt.com/g/g-ssONeKOmx-image-clone-genie)" to play around with?

> https://chatgpt.com/g/g-ssONeKOmx-image-clone-genie

Next time you see a cool image, upload it and see how it describes it. The prompt it gives you can be a starting point, but you'll need to iterate and tweak it to get what you really want.

Here's a pro tip — think about the essence of AI-generated art:

> AI is best at creating things that don't exist in reality.

So, let your imagination soar and think about things "you can't see with the naked eye" — like mythical creatures, impossible architecture, dreamscapes, ancient high-tech gadgets, and more.

If you ever feel your imagination running dry, you know what to do, right?

> I want to explore ideas that don't exist in reality and have DALL·E bring them to life. Please generate a wide variety of creative and imaginative one-sentence concepts, and list as many as possible.
>
> https://chatgpt.com/share/670e4cf5-8b10-8009-89c9-e8a14e6aa770

12.4.2. Fiction Writing: A Classic Example

When it comes to writing fiction, there's one book you just can't skip — Robert McKee's "Story: Substance, Structure, Style and the Principles of Screenwriting." It's a classic.

Robert McKee is hailed as one of the most influential screenwriting gurus in the film industry. His "Story Seminar" has drawn countless filmmakers, writers, and screenwriters, including big names like Peter Jackson and Quentin Tarantino, along with many Oscar winners. His work and teachings have left a mark not just on film and TV, but also in advertising, gaming, and novels.

This book is the kind where skimming the summary just won't cut it. You need to dive in, read it thoroughly, and revisit it often. I promise, even if you never write a story, you'll gain a ton from this book. At the very least, you'll develop a "professional critic's eye" for movies and novels. Plus, the truly valuable insights are those that can be applied across different fields. The real world is interconnected, with similarities across various domains. Even when I'm teaching or writing non-fiction, "Story" offers invaluable guidance.

You can upload the text of this book into ChatGPT or NotebookLM and ask all sorts of questions. Here's a question I highly recommend:

> I want to write stories. Based on the principles from "Story: Substance, Structure, Style and the Principles of Screenwriting, provide a detailed list of various GPT assistants I can create using ChatGPT, with each assistant responsible for a specific task. Clearly describe the specific function of each assistant and explain how these assistants can work together to help me write more efficiently, optimize the story structure, and improve the overall quality.

You don't even need to upload the whole book. Just ask — this book is such a classic, it's been around for years, reprinted and discussed extensively online. ChatGPT knows it inside out. Honestly, its response is pure gold.

> Based on the principles from **Robert McKee’s** *"Story: Substance, Structure, Style and the Principles of Screenwriting"*, here is a detailed list of **various GPT assistants** you can create using ChatGPT to help you write stories more efficiently, optimize your story structure, and improve the overall quality.
>
> ### 1. Plot Architect Assistant
>
> **Function**: Helps develop the **three-act structure** (setup, confrontation, resolution), plot points, and story beats, ensuring that the narrative arc stays compelling.
>
> **Responsibilities**:
>
> - Brainstorm central conflict and turning points.
> - Break the story into key milestones (inciting incident, midpoint, climax, etc.).
> - Ensure that each scene or beat aligns with the overall narrative arc.
>
> **How it helps**: It ensures that your story has a solid structure, preventing plot holes and helping you maintain tension and pacing throughout the narrative.
>
> ------
>
> ### 2. Character Development Assistant
>
> **Function**: Focuses on building multi-dimensional, compelling characters with clear motivations, backstories, and character arcs.
>
> **Responsibilities**:
>
> - Develop character bios, personality traits, and motivations.
> - Map out character arcs and how they evolve throughout the story.
> - Analyze how character desires and conflicts drive the plot.
> - Suggest character traits that would create tension or contrast between characters.
>
> **How it helps**: It ensures your characters are dynamic, believable, and essential to driving the story forward. This assistant can provide suggestions for dialogue and character-driven scenes.
>
> ------
>
> ### 3. Theme Consultant Assistant
>
> **Function**: Keeps track of the **central theme** and sub-themes, making sure that your story delivers a cohesive message.
>
> **Responsibilities**:
>
> - Identify the core theme (e.g., love, revenge, justice).
> - Weave the theme into key moments of the story.
> - Ensure that sub-themes resonate with the larger narrative.
> - Point out where the theme may be inconsistent or unclear.
>
> **How it helps**: Helps maintain focus on the deeper meaning of your story, ensuring that the themes resonate through subtext, symbolism, and key plot events.
>
> ------
>
> ### 4. Scene Analyzer Assistant
>
> **Function**: Examines individual scenes for effectiveness, dramatic impact, and purpose.
>
> **Responsibilities**:
>
> - Evaluate each scene’s contribution to the overall plot and character development.
> - Suggest how to heighten conflict or tension within a scene.
> - Check if the scene introduces new information or advances the plot.
> - Ensure that scenes are visually and emotionally engaging.
>
> **How it helps**: It ensures that every scene matters and enhances the dramatic tension, pacing, and emotional engagement of the story.
>
> ------
>
> ### 5. Dialogue Coach Assistant
>
> **Function**: Specializes in **natural, character-specific dialogue** that reflects character personalities, motivations, and conflicts.
>
> **Responsibilities**:
>
> - Craft character-specific dialogues that reflect personalities and intentions.
> - Ensure that dialogue reveals subtext, avoiding direct exposition.
> - Suggest variations in speech patterns for different characters.
> - Analyze and revise conversations to increase dramatic tension or emotional depth.
>
> **How it helps**: Provides dialogue that is engaging, authentic, and reveals underlying conflicts or emotions, which can elevate the character interactions.
>
> ------
>
> ### 6. Structure & Pacing Assistant
>
> **Function**: Focuses on ensuring proper pacing, act transitions, and story rhythm.
>
> **Responsibilities**:
>
> - Review the pacing of the story across acts, scenes, and beats.
> - Analyze act transitions and make suggestions for smooth flow.
> - Ensure that the climax and resolution are appropriately timed.
> - Suggest when to quicken or slow down the narrative for tension or breathing room.
>
> **How it helps**: It ensures that your story flows smoothly and keeps the audience engaged, preventing uneven pacing that could drag or feel rushed.
>
> ------
>
> ### 7. Conflict & Tension Specialist Assistant
>
> **Function**: Focuses on the primary conflicts within the story, both internal (within characters) and external (situational).
>
> **Responsibilities**:
>
> - Identify sources of conflict and tension.
> - Suggest ways to escalate conflict throughout the story.
> - Ensure that stakes are high and constantly rising.
> - Balance external conflicts (plot-related) and internal conflicts (character-driven).
>
> **How it helps**: Ensures that the tension and conflict remain at the core of your story, keeping readers emotionally invested and driving the plot forward.
>
> ------
>
> ### 8. World-Building Expert Assistant
>
> **Function**: Develops the story’s setting, rules, and world details (especially useful for fantasy, science fiction, and historical fiction).
>
> **Responsibilities**:
>
> - Create detailed descriptions of the world’s environment, culture, and history.
> - Ensure that the world’s rules (political systems, magic, technology) are consistent.
> - Help create the ambiance and sensory details for immersive settings.
> - Suggest world-building elements that impact character and plot (e.g., societal restrictions).
>
> **How it helps**: It makes the story’s setting immersive and believable, helping readers feel grounded in the world of your story.
>
> ------
>
> ### 9. Subplot Coordinator Assistant
>
> **Function**: Manages **subplots** to ensure they complement the main plot and add depth without distracting.
>
> **Responsibilities**:
>
> - Suggest subplots that explore secondary themes or characters.
> - Ensure that subplots intertwine with the main narrative.
> - Track how subplots advance or resolve alongside the main plot.
> - Avoid unnecessary subplots that don’t serve the story.
>
> **How it helps**: It adds layers of complexity to your story without derailing the main narrative, enriching the overall experience.
>
> ------
>
> ### 10. Editing & Style Assistant
>
> **Function**: Provides a focus on **language, tone, and writing style**, ensuring clarity and consistency.
>
> **Responsibilities**:
>
> - Offer suggestions for sentence structure, clarity, and conciseness.
> - Ensure that tone and style are consistent with the story’s genre.
> - Identify repetitive phrasing or unnecessary exposition.
> - Make suggestions to enhance the emotional impact or descriptive language.
>
> **How it helps**: Ensures that your prose is polished and aligned with the emotional and thematic tone of the story, improving readability and engagement.
>
> ------
>
> ### 11. Scene Sequencing Assistant
>
> **Function**: Helps with **rearranging scenes** to improve flow, suspense, or emotional engagement.
>
> **Responsibilities**:
>
> - Review the order of scenes and suggest rearrangements for greater impact.
> - Ensure that each scene builds logically on the previous one.
> - Use flashbacks or nonlinear storytelling where appropriate.
> - Balance quieter character moments with action-heavy scenes.
>
> **How it helps**: Helps you build a more engaging narrative flow, with strong transitions and logical scene progressions.
>
> ------
>
> ### 12. Symbolism & Imagery Assistant
>
> **Function**: Focuses on identifying and enhancing **symbolism, motifs, and visual imagery** in your story.
>
> **Responsibilities**:
>
> - Suggest symbolism that reflects the theme or character development.
> - Incorporate recurring motifs that reinforce the central message.
> - Help craft vivid imagery that engages readers’ senses and emotions.
> - Analyze whether the story uses visual metaphors effectively.
>
> **How it helps**: Adds depth and literary richness by ensuring your story uses symbolic elements and striking imagery.
>
> ------
>
> ### How these assistants work together:
>
> These assistants work as an interconnected team, ensuring that no element of storytelling is neglected. For example:
>
> - **The Plot Architect Assistant** can lay down the structure while the **Subplot Coordinator Assistant** enriches it with subplots.
> - **The Character Development Assistant** and **Dialogue Coach Assistant** collaborate to ensure that character motivations and personalities are clearly expressed in speech and actions.
> - **The Conflict & Tension Specialist Assistant** works closely with the **Scene Analyzer Assistant** to make sure every scene drives the conflict forward.
> - **The Theme Consultant Assistant** ensures that all assistants stay aligned with the overarching theme.
>
> Each assistant has a focused responsibility but works synergistically to create a cohesive, dynamic, and polished story. By delegating specific tasks to each assistant, you streamline the writing process, keep an eye on important elements (like pacing, conflict, and character arcs), and ultimately create a richer, more compelling narrative.

By now, you're basically a "GPT creation wizard." At the very least, you're someone with a "GPT Builder" badge. Once you've read this book thoroughly — and I mean really digested it — you can start iterating and optimizing to build your own "bot army" for storytelling. It's going to be a blast, and you'll probably get so into it that you'll lose track of time.

But here's the thing: bots and most people have a lot in common, especially when it comes to teamwork. So, as the "commander of your bot army," it's crucial — actually, it's a must — to keep these points in mind:

> * You're the leader; they're the support.
> * You do the thinking; they do the grunt work.
> * Without your guidance, they get lost.
> * Without your strength, they falter.
> * Only you care about surprises.
> * Only you are accountable for the outcomes.

At the end of the day, from every angle and at every level, no matter how you slice it, bots won't take any responsibility. Blaming the bots just highlights your own shortcomings.

12.5. It's All About Language Arts

I've lost count of how many times I've said this:

> What's left is just **middle school-level reading comprehension**...

Maybe "language arts" is the most packed yet underrated subject in school — because nobody thinks they're not proficient in their native tongue. In the past, this wasn't a big deal. But suddenly, with the rise of AI, not taking language arts seriously has become a huge, glaring issue — because AI can now directly use natural language.

Humanity's rapid progress is built on the foundation of written language. Without it, we'd be like other animals, evolving slowly and randomly through genetics and mutation. But with writing, we can communicate instantly, across regions and eras, and record, share, think, and converse in increasingly complex ways.

It's no exaggeration to say that all civilization is built on the written word.

Every improvement in writing tools, every standardization of language, every revolution in communication technology has sparked massive global progress. The more people who master language, the more it's shared, the more ideas clash, and the more innovation happens. According to Metcalfe's Law, the value of a network is proportional to the square of the number of its nodes. That's why humanity's growth curve keeps getting steeper.

Today, we not only have language, but we also have precise distinctions: the flexible yet sometimes ambiguous **natural language**, and the simple, precise, unambiguous **artificial language** used in science and tech fields like math, physics, chemistry, logic, and programming.

Some folks got it twisted, dividing knowledge into arts and sciences as if they're two opposing camps. But to the brain, it's all language arts. If there's any difference, it's just that one might use more natural language, while the other leans on artificial language. The catch is, you can't complete intricate tasks with just natural language, and you can't spread ideas widely with only artificial language.

So, no matter what knowledge or skills you're learning, to the brain, it's all language arts — even those skills that seem purely physical. The brain needs to understand, summarize, categorize, review, search, access, and reflect. Most importantly, it needs to communicate — all of which rely on language.

Many parents haven't caught on yet. It's not about "my kid doesn't have a knack for math." In 99% of cases, struggling with math is due to not grasping language arts — poor reading comprehension. The problem isn't the math itself; it's that the question is right there in black and white, but it wasn't fully read or understood. A condition was missed, or another possibility wasn't considered, leading to mistakes.

You need to be someone who values language arts, whether it's your native language or a foreign one. If you're a parent, you should influence your kids to prioritize language — because "reading comprehension" is the foundation of all "basic skills." Without it, you can't learn anything well, or at all, let alone "produce" or "create."

But here's the kicker: it takes a lot of time and effort.

For native language arts, you might have 12 years of daily classes in school, and many still don't master it. Proficient English reading skills often take another 4 years of college and still aren't fully grasped by many. Now, with AI's help, what used to take decades might be shortened, but it still usually takes at least two to three years. Patience is key. Without it, well... fill in the blank yourself.

12.6. Tweaking Your Brain's Safety Threshold

There's a ton of theories and research on why kids seem to learn faster and better. The most convincing one, in my words, is this:

> For kids, it's not just learning — it's a **struggle for survival**.

Our brains are obsessed with saving energy and terrified of dying. That's because the brain's main job is to keep the body safe. After all, the brain only matters if the rest of the body is functioning.

So, your brain won't let you work beyond your limits. Just like a car engine, if it runs at 100% capacity, it won't last long before it breaks down. Engineers set a safety threshold, like capping the engine's power at 70% of its max. This keeps things safe and stable over the long haul.

Why do kids seem to "learn everything"? Because from their perspective, everything is essential for survival. If adults can do it, they must too. Simple as that. At that age, their brains set high thresholds for everything, almost always aiming for 100%.

Why do we lose that drive as we grow up? Because our brains have repeatedly confirmed, "Even if I don't get this, life goes on. I can eat, sleep, and survive just fine." So, the brain lowers the safety threshold from 90% or 85% to 60% or 50%, maybe even less.

Often, whether someone learns or masters something has nothing to do with IQ. If there's any cause-and-effect relationship, it's the opposite of what many think. IQ isn't the "cause," it's the "effect." When the brain sets a low threshold, you quit halfway, and the result is your "IQ" doesn't improve, or relatively speaking, it drops.

Why do so many people read tons of books, hear loads of advice, yet take no action or change? It's tied to the safety threshold their brain "secretly" sets — and they're clueless about it. This ignorance leads to anxiety, leaving them in a corner, pulling their hair, asking, "What's wrong with me?!"

The evolutionary gains from hundreds of thousands of years of genetic inheritance aren't suited for this era of information overload. The brain is a network of neurons, more complex than the galaxy, while knowledge is an ever-expanding network. The world is constantly changing, with trends accelerating and growing. Honestly, the brain struggles to keep up with these changes and their speed.

The least adaptable part might be those default safety thresholds. For hundreds of thousands of years, survival was the issue, so the brain prioritized it above all else. Suddenly, survival isn't a problem. Life is super safe, and the cost of essentials is dropping.

In this scenario, sticking to those ancient safety thresholds means doing nothing — the ultimate "lying flat." Two or three meals a day, a glass of milk, two eggs, and half an apple are enough to survive. Survival isn't an issue and shouldn't be, so why not lie flat?

Now you get it. From a brain science perspective, "laziness" can be clearly defined as:

> Lowering your brain's safety threshold to the minimum...

So, being "proactive" is just the reverse:

> Find ways to raise your brain's safety threshold...

Here's a four-word phrase that can effectively boost your brain's safety threshold when needed:

> **Worse than death**

There are three main templates for using it in a sentence:

> * Can't create..., worse than death.
> * Can't master..., worse than death.
> * Can't learn..., worse than death.

Simple but effective — because our brains really fear death. The brain is fascinating; it can be tricked by anyone, especially by your own self-deception, which is the most direct, effective, and often "instantly effective."

When needed, use these templates, say them to yourself a few times over a day or two. Your brain will remember it deeply, believe it wholeheartedly, and quietly raise that "safety threshold." Naturally, you'll feel more motivated, more resilient, and surprisingly, not tired or annoyed. Everything becomes interesting. Once you experience this magic, you'll never let it go.

12.7. The Ultimate Equalizer

No doubt about it, our analysis leads to a clear conclusion — tools like ChatGPT, representing "generative AI based on large language models," are another "democratized production tool." And not just any tool, but one that can massively impact how we think, produce and create.

Equally undeniable is the fact that, just like in the past, people have often overlooked the power of writing as an equalizing resource and willingly or unknowingly abandoned computers as equalizing tools. It's likely that even more people will, in reality, pass up this incredibly powerful equalizer.

We don't even need to guess. In the blink of an eye, we've already seen this trend unfolding, haven't we? At least from where I'm standing, it's pretty alarming.

The choice is yours.

"Choosing itself is a form of production."

Choice is the last asset anyone has, the ultimate asset — no one can take it away unless you give it up.

Go ahead, make a sentence with "worse than death."

You got this!

You absolutely got this!

Copyright: © LookMinds. Our own edited text and translation, not a substitute for the original; rights in the original belong to the author and the platform. Please see the original link.