Ask AI, how did I do it?

AI conversations are completely different from human ones. They are based more on logic, data, and algorithms, and they focus more on information transmission and problem solving. Therefore, the biggest problem with AI software is not "how difficult is the software to operate", but "how to ask a good question". How should we ask questions to artificial intelligence, how to define the problem, and how to grasp the relationship between variables in the problem? The author of this article analyzes this and hopes it will be helpful to you.

AI conversations are completely different from people.

The human conversation process is based on emotions, cultural background, and current situation, and needs to consider non-verbal factors such as the other party's emotions, tone, and expression. AI conversations are more based on logic, data, and algorithms, and pay more attention to information transmission and problem solving.

Although all companies are promoting natural language processing technology (NLP), and AI dialogue (Wenxinyiyan, chatGPT) is gradually moving towards a "humanized" direction, there are still obvious differences between the two in essence.

Faced with the emergence of AI, many people have a certain level of ideological awareness and are reducing their self-substitutability by learning knowledge. However, they find that the biggest problem is not “how difficult is the software to operate” but “how to ask a good question”.

Kevin Kelly once mentioned 12 major trends for future development in a famous speech, and the 11th trend is asking questions; however, in the current management market in China, people have not yet made full use of asking questions.

What is asking a question? How should we ask questions to artificial intelligence, how to define questions, how to classify and transform questions, how to reduce big questions to small ones, and how to grasp the relationship between variables in the questions? Let me try to explain it clearly.

1. The logic of the question

What is asking a question? Asking a question is to raise a question. The question is the gap between the goal (standard) and the reality.

For example, if the goal is to get 90 points, but the actual score is 60, the gap between 60 and 90 is the problem. How can we raise the problem? We need to be able to "discover" and "identify" the problem first.

Three elements closely related to asking questions are "what, why, and how", and the core of asking questions is to figure out the relationship between the three. In 2009, American marketing consultant Simon Sinek used it for the first time to express the leadership model in his TED speech. This model was later called the Golden Circle Rule.

Let’s first say what it is.

People often get confused here because what usually includes four invisible conditions, namely: definition, connotation and extension, making judgments, and the connection between concepts.

Just like "love", at the definition level it refers to a deep feeling that manifests itself as selfless concern, care and emotional attachment to others. It can also be family affection, and different people have different views and experiences.

But what will this lead to?

We often see a boy who has given a lot but still has no love in the end. The two sides have different standards for defining love and the boundaries are not clear, so it is naturally difficult to create sparks.

What is the answer? We need to pay attention to the relationship between concepts to avoid logical errors. For example, "Yangcheng" and "the capital of Guangdong Province" refer to the same city, Guangzhou, but the two concepts are in the same relationship.

There are six kinds of relationships between concepts: identity relationship, inclusion relationship, intersection relationship, contradiction relationship, opposition relationship, and logical parallel relationship. (Practical Logic Course (5th Edition) by Zhang Mianli, China Renmin University Press, 2015).

It can be seen that even though you know the fact that there is a 30-point difference, you may not know "what" caused the difference. What should you do? Ask "why". Why is a causal relationship, which is the truth behind things.

for example:

Why does an engine work? There are physical principles behind it. Why does an airplane fly? There are aerodynamic principles behind it.

We know that innovation, progress, and problem-solving to achieve results are all very important; so, is there a causal relationship between wanting to innovate, progress, and problem-solve to achieve results and asking questions?

Yes. If you don’t know how to ask questions, you can’t find the problem. If you can’t find the problem, you can’t solve it, and then you can’t get results. This causal relationship also further explains “why asking questions is important”, after all, asking questions can bring us better things.

Besides, how to do it.

After the rain, there was a puddle of water on the road. The child saw the adults stepping over it and tried to do the same. Unfortunately, because his steps were too small, he stepped into the water, which made the adults laugh. The child learned how to do it by observing and imitating.

Therefore, “asking questions” seems to be a solution to the problem, but in fact it is pursuing something better.

Just like when you see an animal and you don’t know what it is, your first question is “what is this”; when someone tells you that this is a horse, you might ask “what can a horse do”; after someone tells you that a horse can be used to ride and carry things, you will understand “what use is a horse to you”. The whole process of asking questions is a detailed process of cognition of things.

However, in reality the problem is difficult to define.

I have many friends who are in the self-media industry. They are very hardworking and always say they want to write about this or that topic when chatting. But when they sit in front of the computer, they think for a long time and barely type "research on scientific marketing" on the screen, and then nothing happens.

Why?

Because problems include four types: broad issues, problems, questions, and topics. Although these words are difficult to distinguish in Chinese, the scope and meaning they cover are very different.

Most of the articles I write are based on questions and topics. Topics and problems must go through a certain thinking process to be transformed into research questions. If I don’t conduct in-depth research in a certain field, my expression will be rather vague.

What is the difference between the three?

Generally speaking, an issue refers to a broad topic , which is a cluster of problems covering a wide range of areas. For example, the metaverse includes AR, VR, content systems, operating systems, digital twins, and other things that you simply cannot understand, or you will be stuck in a situation where you are trying to eat the sky and have no idea where to start.

Problems are real difficulties and troubles that require action and intervention to alleviate, such as the decline in food safety and fertility rates at the macro level, and the inability to find a job (employment), a partner (marriage), depression (psychological), etc. at the bystander level.

There is overlap between problems and issues.

However, it can be big or small. You can ask questions at the middle level, that is, integrate them into the question link, like how to solve the "employment difficulties of college students in the context of large-scale layoffs", and you can relatively find a focused answer; the topic is a discussion of a specific event or phenomenon; just like Coco Tree frequently appears in the circle, everyone holds different opinions, and so on.

However, many times the problem you face is how to add a question mark to a declarative event after obtaining certain background information, which involves mastering the types of questions asked.

2. Questioning Methods

Question type? Isn't it just a direct question?

For example, notion AI, help me answer, what are the difficulties of Xiaohongshu in doing e-commerce? Chat-GPT, help me write an outline for an article about scientific diet.

It has been proven that AI is better at solving two types of problems: one type is clearly structured , such as writing code or doing a routine task; the other type has a clear framework , which only requires it to fill in the details, and you have clear requirements for the details. Therefore, open problems may have no solution in themselves and do not have much practical significance.

What to do? You can try this method.

First, closed questions

It mainly answers factual questions, using who, what, and when. It is like filling in the blanks, often with a definite answer, but once the answer is given, it ends without much extension.

for example:

When I'm looking for various concepts, I'll ask chat-GPT to help me find out what the Golden Circle Principle is, who proposed it, what its significance is, why the proposer thought so at the time, and give as many examples as possible.

Or, what are some of Charlie Thomas Munger's speeches on critical thinking? In what scenarios, what kind of key role do these speeches play?

Second, why and how

When using analytical questions, I would use “why and what”, which can go beyond the purely factual level and connect the dots to find order in chaos; for example: I asked AI Douyin why it would do food delivery, and what would happen after it did it? The answer was as follows:

I can only blame myself for asking too shallow questions. This answer may not be satisfactory. How can I shift the focus from "broad topics" or "real problems with only background" to deeper questions?

I propose a three-step fill-in-the-blank questioning method:

First step, I have to research;

In the second step, specifically, ① I want to focus on the following questions: Why do some... and others...? (Here, compare the differences in the phenomenon) ② What influenced this result, ③ What is the mechanism between these factors and the results, and how is it?

The third step is to solve practical problems after answering the above questions.

In the first two steps, we distinguish between issues, problems, and questions, respectively; this way, we can start from a broad topic and focus on a few issues for research. The final answers can help us solve real problems and advance our arguments.

The second step involves:

Why (why), discover interesting contrasts, paradoxes and differences from the real world; what factors (what), make bold assumptions and look for possible factors that affect the results; how (how), carefully verify and clarify the role between principles and results.

For example:

I want to study (why Douyin wants to do food delivery). Specifically, I want to focus on the following questions: ①Why does Douyin want to do food delivery, and where did it start? ②What factors will affect the expansion of Douyin's food delivery? Some say it is about the riders? Some say it is about the merchants. ③What is the mechanism of these influencing factors?

Of course, AI’s answer can only serve as a guide, and the answer it gives may be relatively broad. After all, it is based on the “existing massive content” as an integration. If you want to understand in depth, you still need to conduct real investigation and expansion on the details.

For example, how many riders there are in Chaoyang, Beijing, how many orders they deliver every day, how much money each rider earns, how long they work, etc.

Quality control circle (QCC), the epitome of Japanese quality management, Ishikawa Kaoru proposed the "five whys" method, which I also often use.

The specific method is:

Identify a problem, ask why questions for the problem, find the answer and ask why again; repeat steps 2 and 3 until you reach the root cause, find the root cause and solve it.

Simply put, this method explores the problem in depth by repeatedly asking "why" and then finding "how to do it", which seems a bit like the first principles that Elon Musk often talks about.

I think only a few people do this. After all, questioning is more about "writing, copywriting, improving sentence structure, and improving the three-dimensional sense of expression". Unless you are doing in-depth reporting and need to dig into the essence, otherwise, you can't go so deep.

How do I ask for details?

It’s not about asking “AI, why do you do this?”; instead, it starts from the six elements (who, what, when, where, why, and how). I often use who, where, when, and what to continuously close the gap and achieve effective focus.

For example:

Let’s take the psychological research of college students as an example. First, we need to narrow the scope of the research subjects (who) from Chinese college students to undergraduates, or to a few majors.

Then, we can narrow the time range (when) and limit the research subjects to a certain grade, such as freshmen, and further refine it to the first semester of freshmen.

Thirdly, narrow the geographical scope (where). As above, narrow it down to a certain region, such as Beijing or Shanghai. In this way, the geographical scope of the research will be very concentrated.

Finally, you can also narrow down the research topic (what). Psychological problems are very complex and diverse, and each symptom is different. If you focus on one or two, the problem will become more controllable, such as: research anxiety.

What do we get from this operation? China, Beijing, freshman, first semester, student anxiety.

The reason for doing this is that, on the one hand, the information provided by AI is too vast. If the time span of your inquiry is large and the scope is unclear, it will be drowned in the vast sea of ​​historical materials. On the other hand, vague questions can be transformed into specific and controllable research by narrowing the focus and adding qualifiers.

Some people may say that if we narrow the problem down to a very specific and narrow angle, isn’t it trivial?

Like me, if you use AI to improve the efficiency of writing and business research, then we are more concerned about "one-sided and profound" rather than "comprehensive and superficial."

Assume two situations. One is to keep working on the so-called "big issues" (such as headlines, Alibaba business analysis, industry analysis, and competition landscape analysis). Since the issues are too complex and cannot be effectively divided, they may end up being written into some irrelevant empty words and clichés, the so-called "big pie articles."

Another approach is to focus on a specific issue in the competitive landscape and ignore the rest. Then, you can see the big picture from the small details of a big problem. It is like playing chess, focusing on the big picture but not the small details. This is an advisable path.

When it comes to writing and business research, the importance of asking questions is unquestionable. The three-step questioning and five whys are worth your deliberate practice. Give it a try. May you become a “problem youth”.

However, when you first use AI, you may find the “serial questioning method” too complicated, so you will directly give it some “simple problems”, but you don’t want to get “vague” answers.

What should I do? The sentence pattern “some…some…” mentioned in the previous process may be more suitable for you.

3. Question Variables

Let me tell you a story first. Many phenomena in our daily lives are very strange, but we often turn a blind eye to them.

Have you ever been to a bar? Have you ever noticed the height of the bar stools? About 70cm, while the height of the stools in daily life is 40-50cm. I know what you are thinking: Why are bar stools so much higher?

The height of bar stools and ordinary stools forms an interesting contrast and a contradiction to common sense, hence the question mark.

If I tell you that a certain "health master" died at the age of 51, you will definitely be shocked. I call this kind of problem a puzzle. Simply put, the inconsistency or contradiction between new and old facts will make people feel uncertain, leading to confusion or curiosity, so that they want to solve the problem.

The puzzle of the question is here.

Steven Levitt cited many examples in "Freakonomics", for example, drug dealers seem to make a lot of money, so why do they still live with their elderly mothers?

For example, why do some parents often read to their children and often take their children to museums, but their children are not successful? Why do those who don’t read much and don’t go to museums often succeed? These questions go against intuition and quietly plant a question mark in my mind.

How to design a puzzle?

Take a question as an example, which group do you think is more numerous in China, doctors or nurses? Most people would think nurses, because it is common sense; but if I tell you that "the number of doctors in China has long been greater than that of nurses, and only in recent years have nurses surpassed doctors", what would your reaction be after hearing this fact?

If your first reaction is "why", then it is a good question. If you answer "oh" after listening to it, it means that the question is unremarkable. Find those phenomena that are different from common sense, reality or theory, and find those problems that make people puzzled and feel puzzling, you can get different answers from AI.

For example:

Trains, planes and other means of transportation can save a lot of time, so why do people always ask where the world has gone? At this time, AI will tell you that we must be clear about our goals and ideals, and then use transportation tools to achieve them.

If we ask directly, "Where does time go?" or use "Where does time go" as a proposition to discuss the importance of time, you can only get a bunch of "what"s, which may not be of much use to you.

A classic example comes from the skit "Selling Crutch" by Zhao Benshan, Gao Xiumin and Fan Wei. There is a classic line in it: "I wonder how can a couple who live together be so different in their conduct?" This line is well known in most of China.

But what everyone doesn’t know is that it shows an interesting way of asking questions. This sentence is half control and half comparison. It controls the life situations of two people and compares the differences between the two people; thus forming an interesting puzzle.

So, where do these weird angles come from?

After a lot of conversations with AI, I think variables are very important, that is, after you give the background in the beginning, the two questions behind form a contrast. With the sentence pattern of "some...some...", you can successfully transform the puzzle into a question.

Why do some places… some places…? Why do some people… some people…? Why do some times… some times…? We can gain new perspectives by comparing the differences in variables across regions, individuals, and time.

This sentence actually concretizes the three question marks of where, who, and when.

Those who are not good at asking questions can consciously train their questioning skills by using “some…some…”. Although it seems a bit mechanical, this sentence pattern is like a baby walker, which can help beginners quickly master AI dialogue skills.

for example:

Why do some people in Internet companies shout for compensation when they are laid off, while others remain silent? Why do some children growing up in the same family behave well while others are lawless?

Why are some people happy while others depressed when they live the same life? Why do some people still have thick hair while others have thin hair when they stay up late? Why is red wine more elegant than beer when they drink the same wine? Why are Meituan and ByteDance calm when they are laying off employees while JD.com is always criticized by netizens?

The puzzle lies in the variables. Variables are divided into dependent variables and independent variables. Assuming you want to get a result to be explained, start from the result and try to infer what causes this situation.

Learn to embed this series of methods into your own questions, and then throw them to AI. You can get surprising answers. Don’t believe it? Try it.

IV. Conclusion

If you design the why well, the what and how will not be wrong. When asking why, try the "five whys and three-step questioning method".

If you find it too difficult, start with “some…some…”

Remember to use half to control and half to compare; AI is always a tool. Only by mastering the method of asking questions by AI can you improve the efficiency of problem solving. I wish you can learn and apply it flexibly.

Author: Wang Zhiyuan, public account: Wang Zhiyuan