Last Saturday, I participated in an offline activity organized by Teacher Li Jigang.

The event was full of useful information, especially on topics such as the value of prompts in the AI ​​era and the art of communicating with AI. Teacher Li’s sharing was like an epiphany, giving me a deeper understanding of AI human-machine collaboration.

I have sorted out some of the core gains so that we can think and improve together.

1. AI is becoming more and more powerful, is there still a future for prompt words?

To sum up the essence of the prompt in one sentence: Your "input" affects your "output"

Faced with the surging wave of AI such as Manus, GPT-5, GPT-6, etc., many people may ask: Is it still worth learning prompt words?

Li Jigang's words enlightened me: the essence of prompt words is that your "input" will affect your "output".

No matter what form AI evolves into, as long as we still need to input information (instructions, questions, materials, etc.) into it to get the expected response, then this "input" skill, or "prompt words" in a broad sense, will always be valuable.

The difference may be that as AI capabilities improve exponentially (IQ, EQ, comprehension, etc.), we may no longer need to write out long, detailed prompts like we sometimes do now.

The input may become more refined, but it is no less important.

2. When should we speak less and when should we speak more?

As AI becomes smarter, the prompt words we input may become shorter. So when should we say less and when should we say more?

This requires us to bring out our "old friend", the classic communication model: the Johari Window Model.

This model is used for communication between people, and of course it can also be used for communication between people and AI.

1. Public area: You know, AI also knows - the less you say, the better 

When the content of communication is in the "public zone" known to both humans and AI, you can say as little as possible.

For example, if you want AI to play the role of a philosopher and discuss philosophy with you, just say "You are a philosopher".

No need to elaborate on “you need to understand existentialism, logical positivism, be able to think critically…” and so on.

Why? Because the word “philosopher” has been associated with a vast amount of background information, thinking patterns, and knowledge fields in the vast knowledge base of AI.

You say "philosopher" and it understands.

Another example: imitating writing style.

If you say "write a paragraph in the style of Lu Xun", AI can immediately mobilize all the data about Lu Xun's life, works, and language characteristics (sharp, critical, sense of the times, etc.).

There is no need for you to explain "what is Lu Xun style" anymore.

In the public area, “say less” because the information density contained in the keywords themselves is high enough and both parties (humans and AI ) have a consensus. 

2. Hidden Area: You know, AI doesn’t know—need to say more 

When you need to express some personalized or private information that is unlikely to be in the AI ​​training data, you need to "say more".

For example, I want AI to imitate "Suyuan's expression style" to write articles.

Does AI know who Suyuan is? Does AI know what my style is? Most likely, it doesn’t.

So what should we do?  At this time, we need to "speak more".

Describe my style in detail (for example: I like to use metaphors, strict structure, a little humor?), and even give specific examples (few-shot examples).

The process of “saying more” is like giving AI a “knowledge patch” to help it understand concepts that it originally did not know.  

Providing examples is especially effective because sometimes we ourselves have trouble describing a feeling or style in words, so it’s better to just show it: “Look, it’s like this, you figure it out for yourself!”

3. Blind spot: You don’t know, but AI knows---test your questioning skills 

This is an area that AI knows about but we don't.

In the AI ​​era, this is precisely where the huge opportunity lies. When we have AI, a powerful "external brain", as long as you want to learn and know how to ask effective questions, you can pull this lever to quickly acquire and learn a massive amount of knowledge and information that you did not know before.

4. Unknown Zone: You don’t know, and neither does AI —just chat 

This is the common knowledge boundary between humans and AI.

Li Jigang said that he didn't know much about this field, but I personally think that this is where curiosity and the spirit of exploration come into play, and we can chat about it casually whenever we want.

Perhaps, through free conversation and collision with AI, we can really inspire inspiration and reach unknown areas together.

3. Why should we “speak less”?

Since “saying more” can help AI understand my intentions more accurately, why not just describe all situations as detailed as possible?

This is because overly specific instructions may limit the AI's room for maneuver and sacrifice the breadth, depth and even creativity of the results.

For example, if I add “existentialism” in the prompt of “philosopher”, the AI’s answer will most likely be framed within the scope of existentialism, and miss the possibility of bringing surprises from the perspective of other philosophical schools.

Therefore, under the premise that AI can understand, "saying less" can sometimes inspire AI to perform better.  

4. AI will not bring about equal rights in knowledge, but may instead exacerbate divisions

We talked about how AI can be a powerful lever for knowledge, but does this mean that everyone can easily access knowledge and achieve “knowledge equality”?

Li Jigang's point of view (which I strongly agree with) may pour cold water on you: AI will not only not bring about equal rights in knowledge, but may instead exacerbate the knowledge gap between the rich and the poor. 

The reason is simple:

The willingness to actively learn has become divided. In an environment of instant gratification and dopamine stimulation such as short videos and information streams, people's concentration continues to decline.

The number of people who are truly willing and able to calm down and actively use AI for deep learning and thinking is destined to be a minority.

The low completion rate of paid courses on platforms such as Bilibili shows that for many people, "learning" is just a psychological comfort when purchasing.

If you don't carry out systematic learning for a long time, your learning ability will be wasted.

The result is that only a few people who know how to collaborate with AI and maintain an active learning attitude can use AI as a super lever to achieve a leap in knowledge and ability and become the "rich" in the information age.  

5. Become the minority who can control AI

Deep insights into how to learn, how to think, and how to stay ahead in the AI ​​era.

The key is to continuously improve the ability to communicate with AI efficiently and maintain the habit of active learning and deep thinking.  

As for how to truly become one of the "few people"? I have also been exploring and practicing recently. I have some ideas, but they are not mature yet. I will share them with you when the time is right.

In the era of dancing with AI , challenges and opportunities coexist. May we all be the one who holds the reins and controls the horse.  

(The illustrations in this article are all generated by the latest GPT-4o image model, which transforms the scene of the event into the style of Ghibli. The effect is very good~)

This is Senior Suyuan, a post-00s who doesn’t want to compete with others and is using AI to build a healthy and free lifestyle.

Committed to using "AI + system thinking" to help 1,000 people build three major systems:

- Information input system: filtering noise and accumulating knowledge

- Content output system: stable output and long-term monetization

- Personal growth system: focus on habits, sustainable rhythm

I don’t like to follow the trend. I only share AI tools, AI workflows, and growth records that I have personally used and are truly helpful in my life and work.