How to get used to AI

I let ChatGPT (GPT4) play the role of “questioner unfamiliar with AI” and answered the question. On my side was iPhone voice input. https://chat.openai.com/share/494edcb4-04eb-483d-a6dc-e3f01cda4702

• This itself is a demonstration of “let AI ask questions”, “let AI play a role”, “let AI generate ideas”, and “let AI replace time-consuming experiments”.

Extract the part of the connection as a Q&A in the dialogue. The following is an extract of the connected parts of the dialogue as Q&A.

Q: What methods do you prefer to use to learn about AI technology? Please describe your specific learning style - online courses, books, hands-on projects, etc. A: When it comes to getting information about AI technology, this is a personal story for me, but I am a person who reads papers, and I find it very beneficial to have AI assist me in reading papers. For the general public who will be listening to this talk, I don’t think they will be learning AI by reading papers, so I think this explanation is not very good. For example, there are some government documents that contain a huge amount of important information, but it is difficult to read all of them.

Q: When verifying information obtained from AI, are there any particular points that you focus on? For example, do you focus on the reliability of the source of the information, the freshness of the information, or the consistency of the information? A: When it comes to accuracy of information, you should take different paths to find out what you want to be accurate. For example, if the information obtained through different paths, such as asking an AI a question and then doing a Google search on the keywords that came up and looking at Wikipedia, looking at books, or asking people, is consistent and connected with consistency, then it can be considered highly reliable! Ah, in terms of verifying that. Oh, in terms of verifying the information, one of the most important options is to verify it by experimenting on your own.

Q: Are there any particular uses or innovations that you find particularly effective in using AI to gain new ideas and perspectives? For example, how to ask certain questions or how to proceed in a dialogue with the AI. A: One way to use AI to gain new perspectives that I find effective is to specify the AI’s role. For example, when I think about something, my perspective is inevitably biased and biased toward the perspective of a human being who is a programmer, a researcher, and loves technology. Therefore, when I ask AI a question about some topic, if I assign a role, for example, “You are a non-engineer who is not good at technology,” then it is more likely to come up with statements from a perspective that I did not have.

Q: After receiving a response from an AI with a role designation, how do you incorporate that information and ideas into your own thought process and projects? I would appreciate specific examples. A: As a specific example, I think this is itself what I am doing now. In other words, since I personally am an AI expert, it was difficult for me to think about what kind of questions I myself would ask from someone who is not familiar with AI. Therefore, I gave the AI a role to ask questions, and specified a character who is not a researcher who reads papers, but an ordinary user who is going to learn AI.

I think you’re missing some extraction, especially with regard to questions that might come from novice users, so try again. Q: Are there any learning resources or books you would recommend for beginners to start learning AI? A: The first thing I avoid is social networking sites, which are very biased because they are prone to a barrage of excessive and radical statements. I would trust what my friends tell me that they have tried this in closed groups or chatting with friends who know each other face to face, and I would trust what my friends tell me that they have tried and liked. Also, it is very important that it is something that you have experienced firsthand, not just information sent out by others, or text, or on the Internet. I believe that one must see with one’s own eyes.

Explain an interesting point. The conversation presented includes several interesting points about the use of AI and learning approaches. In particular, the following points are worth noting

• Approaches to Efficiently Reading Information Using AI: Methods for efficiently extracting necessary information from articles and large amounts of material using AI are discussed. Specifically, it is proposed to use AI summarization and answers to questions to narrow down the amount of information to be read and control the depth of reading according to the purpose. This can be an effective means of increasing productivity in this age of information overload.
• Approaches to Using AI to Gain Different Perspectives: This section introduces the use of assigning a specific role or character to an AI to elicit opinions and ideas from perspectives different from your own. This is a creative way to go beyond one’s own knowledge and biases to gain new insights and ideas. Incorporating diverse perspectives will enhance the quality of problem solving and decision making.
• Approaches to Using AI to Streamline the Product Design Improvement Cycle: In the field of product design, an idea is presented to streamline the design improvement cycle by utilizing AI to act as a user persona and collect and analyze feedback. This is an innovative approach to quickly understand the diverse needs and reactions of users and improve designs accordingly.
• Importance of learning about user diversity through AI: When using AI to collect feedback from users, it is important to recognize that user responses are diverse and sometimes contradictory, and to view this as a learning opportunity. This is valuable experience for developing user-centered design thinking and developing better products. These approaches and ideas point in the direction of using AI not as a mere tool, but as a partner for extending thinking and creativity. This way of utilizing AI will contribute not only to increased productivity, but also to improving the quality of innovation and problem solving.

I think there were some interesting stories in the first half. Extracted from the first half. Yes, there were some interesting points in the first half of the report, which I will extract and discuss.

• Need for correct understanding of AI’s capabilities and limitations: It has been pointed out that excessive expectations and misunderstandings of AI can lead to disappointment and wasted time. in order to understand AI’s true capabilities and make appropriate use of it, it is important to correctly recognize its strengths and weaknesses. In addition, we must be willing to critically examine the output of AI, rather than relying on it.
• Use of Multiple Sources to Verify Information Accuracy: The importance of using different sources of information to compare and verify information obtained from AI is emphasized. When information from multiple sources is consistent, the reliability of that information increases. This is an effective approach not only for AI, but for all types of information.
• AI can be used to test hypotheses and new ideas with less risk and effort. and new ideas, while reducing risk and effort. However, care must be taken in interpreting the results.
• Importance of viewing AI as a trigger rather than a “correct answer”: The importance of viewing AI output not as an absolute correct answer, but as a trigger for new insights and ideas has been pointed out. creative problem solving becomes possible. These points are important perspectives from which to effectively utilize AI and maximize its benefits; a proper understanding of AI’s capabilities and limitations, and using AI as a catalyst for creativity and problem solving while validating information in appropriate ways, will be key to how we interact with AI.

Extract the Pessimistic Misconceptions story. If there are any points that are difficult to understand, please ask questions. The following details about “pessimistic misconceptions” were touched upon in our conversation.

When people try something new, they tend to make a pessimistic mistake based on their initial experience of failure. In other words, they assume that just because it didn’t work the first time, it will not work the same way after that, and they will not try that thing any further.

But in reality, the first failure may have been just a fluke, and yet they fail to recognize that possibility and miss out on valuable learning and growth opportunities.

Therefore, he stated, it is important to be careful not to fall into pessimistic misconceptions and to try again and again without being discouraged, even if one fails.

I think your explanation of this “pessimistic misconception” is generally clear. However, a more detailed explanation of the following points might help us understand it better.

• What are some specific situations or examples where pessimistic misunderstandings are likely to occur?
• Are there specific methods or mindsets that we should keep in mind to avoid pessimistic misconceptions?
• I understand the importance of trying again and again, but on the other hand, I think we also need to give up somewhere in order to avoid wasted effort. How can we strike a balance between trying and giving up?

If you have any thoughts on the above points, please let me know.

Not good example 1

. Description. This text discusses how AI can be used to improve learning and operational efficiency. The main topics covered include.

• When utilizing AI, it is important to understand its capabilities and limitations. Excessive expectations and misunderstandings should be avoided, and AI should be utilized in areas where it excels.
• The summary and question-and-answer functions are utilized as a way to process information efficiently using AI. Clarify the information you want to know, then query the AI to get the appropriate information.
• It is important to verify information obtained from the AI with other sources. Reliability is enhanced when information obtained from multiple sources is consistent.
• In areas where experimentation is easy, such as programming, AI can easily be used for trial and error. On the other hand, AI could be used for simulation in areas where experimentation is difficult.
• AI does not necessarily provide the correct answer. Rather, it should be used to generate new perspectives and ideas. Like brainstorming, it is important to first generate quantity.
• Giving AI a role can elicit opinions from perspectives that you do not have. By assuming different user positions and having the AI ask questions, it can help understand diverse needs.
• The improvement cycle of product development could be streamlined through the use of AI acting out user personas. It has the advantage of simulating a variety of user reactions. The above suggests that understanding the characteristics of AI and using it in the right places can increase the efficiency of learning and problem solving.

Comments: Abstract and not good.

• When utilizing AI, it is important to understand its capabilities and limitations. Excessive expectations and misunderstandings should be avoided, and AI should be utilized where it excels.”
• You don’t know that, so your supposed readers are in trouble!

Not good example 2

. Create reading material for readers with no prerequisite knowledge in the form of Q&A. Q1: What is important to utilize AI?

A1: The most important thing in utilizing AI is to properly understand its capabilities and limitations: AI can outperform humans in certain areas, but it is not good at making common sense or ethical decisions.

Q2: What is the most efficient way to process information using AI?

A2: AI’s summarizing and question-answering functions can help you process vast amounts of information efficiently. For example, if you ask specifically, “Please list the three main points of this document,” AI will extract the main points precisely.

Q3: How can I verify the reliability of information provided by an AI?

A3: It is important to check the information presented by the AI with other sources of information. For example, if you look up the same topic in a web search or in a book, and the contents of multiple sources are consistent, you can judge that the information is reliable. Make it a habit to verify information from various angles, rather than just taking it on faith.

Q4: How can AI help solve problems?

A4: In fields where trial and error is easy, such as programming, AI can be used to quickly find solutions. In contrast, in fields where experimentation is difficult, AI simulations may provide hints for problem solving. using AI to try various approaches may be a shortcut to finding a breakthrough.

Q5: Why should we not be overconfident in AI answers?

A5: AI’s answers are not always correct, and it is wise to use AI’s suggestions only as a reference and as a starting point for new ideas. Just as quantity is more important than quality when generating ideas, let the AI generate as many options as possible. It is the job of humans to find promising ideas among them.

Q6: How can we use AI from the user’s perspective?

A6: One method is to set up an image of a user with a different point of view from your own, and have the AI ask questions from that point of view, e.g., “novice user” or “expert user. Asking questions that take into account the diversity of users will lead to better service development.

Q7: Do you have a message for our readers?

A7: AI is a useful tool, but it is not a panacea. To get along well with AI, it is important to understand its characteristics, not to blindly trust its suggestions, and to think for yourself.

Thoughts: I did this right after Example 1, so it was a watered-down version of the abstract from Example 1.

---

This page is auto-translated from /nishio/AIに慣れる方法 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I’m very happy to spread my thought to non-Japanese readers.