Recently, AI tools have been released one after another. There are many free and paid tools. About a while ago, ChatGPT-4o was released. This is the newest version of ChatGPT. I have watched foreign official videos, which show a variety of very attractive tools. The artificial intelligence function also attracted me to pay to upgrade, use it early and test its various functions. For me, a person who is involved in the field of AI all over the world, this upgrade test is very important because I can know the latest Development, and those deficiencies found can be immediately used in your own research and development site
Two days ago, when I was checking my daughter's math homework in the fifth grade, on a whim, I took a picture of a question with my mobile phone (I asked a simple calculation question) to test whether the new function he released was stable and easy to use, but after the test, I discovered that the answer given to me by ChatGPT-4o was wrong, which gave me goosebumps immediately, because if I just let my daughter directly use ChatGPT to teach her math problems, then I will let my daughter become less good as she learns? I write this article to share this experience with you and give you some suggestions on how to use AI tools more appropriately.
The picture below shows that I was taking a photo of a math problem with my mobile phone and sending it to ChatGPT-4o, asking him to help solve the problem. The question is actually not difficult, because it is a mixture of time subtraction, multiplication, and division. My daughter can solve it by herself, but I helped her confirm whether the answer is correct, because it is an assignment given by the school and she did not have the answer. I helped to identify and teach her the wrong questions she did.
As can be seen from the picture below, ChatGPT-4o is actually very good at scanning, identifying, and recognizing questions. It is clear that it knows the meaning and numbers of the questions, so its function in this area is worthy of praise.
The picture below is his problem-solving process. His design is very good and he will explain step by step how he completes the calculation. This can actually give children a good guide to learning. However, I think his final answer number was not the same with that calculated by me. I was shocked at first. I thought I was so old, or presbyopic. I looked at his answers again, and then I checked my own answers again, but no matter how I checked, the numbers were different from ChatGPT. I originally thought that his misinterpretation was caused by an incorrect interpretation of the question scan, but I took a closer look at his interpretation of the question, and it seemed to be completely correct, so I followed his process carefully, and suddenly found that the addition of three numbers in simple mathematics was wrong by ChatGPT-4o. How could something like this happen? It was just the addition of three numbers, but the result was so wrong. I really couldn't believe it.
Next, I guiltily asked ChatGPT a question, "Are you sure the answer is correct? ", and he responded: "Let me recalculate it again." When after he calculated for the second time, the new answer was actually different from the first time. The answer this time was the same as the answer I calculated. It turned out that I was not old enough to do such simple mathematics with wrong calculation, haha!
At this time, I asked ChatGPT-40 very unhappily: "Why did you give the wrong anser? ", and he responded to me: "I apologize that my answer at the beginning was wrong. Let’s solve the problem correctly step by step.” He was showing the calculation process in detail, and it was still the correct answer (as shown below). I like her attitude towards system design, because she will accept the questions humbly and try to respond.
In fact, I was very happy to see this wrong answer, because I have been studying computers for more than 30 years. I know very well that computer calculations are not absolutely correct, especially for AI systems. I just didn’t expect that I was so lucky to encounter this kind of error at the beginning of using paid AI tools. Haha~~ Due to my experience in solving this error, I would like to suggest some mentality and methods for using AI tools correctly as follows:
- It is better not to believe ALL generated from AI.: The content in the book is not absolutely correct. For example, there are sometimes typos, problems with the discussion, etc. Therefore, the same goes for AI tools. The content may also have errors. Do not completely trust the output of AI tools. After seeing this example, you should take my word for it.
- AI tools are different from writing programs, the answer is not controllable for AI system.: AI tools are different from program systems. Most program systems have controllable and predictable results. Because programs can control logical calculations and processes, the results are less prone to errors. But the AI system is different. The AI system learns and summarizes from a large amount of data through deep learning methods. Therefore, the response to the question is not absolutely controllable and predictable. It depends on how you ask the question and how it learns the material and data, so never completely believe that the AI tool will be correct or always give you a suitable answer.
- Don’t be puppets who only copy answers from AI tools.: Don’t let yourself or your children only rely on AI tools to produce results and use them directly, because when AI tools increase your efficiency, your abilities (organizing, summarizing, thinking, planning, etc.) and IQ will also decrease rapidly. Maybe AI tools can efficiently solve the problem at hand, but in the long run, maybe you will slowly become a tool person. In this case, you will be more and more replaceable in the future. Because other peoples also know and are able to use this tool, what’s the difference between you and them?
- Be able to flexibly interact with AI tools and develop the ability to criticize answers: Because of the third point above, I strongly recommend that you should be the protagonist to complete tasks, and use AI tools to assist you in completing various fragments. You must grow together in the process of using AI tools, so that you can always maintain your uniqueness and irreplaceable abilities.
The above are my personal sincere suggestions, because the more I look at the growth and prosperity of AI tools, the more I worry about the degradation of human beings. Many jobs have been gradually replaced by AI, although some celebrities will say: "Humans just switch to other creative work. will not really be replaced." This sentence is basically correct, but it is only half correct, because in the future you will find more and more idle people, because many people may not even know what they can do. The world will be controlled by a small number of talents (because they want to develop AI systems to replace various jobs), and society will be filled with various robots equipped with AI software. What will humans do in the future? It may be worth thinking about for you and me.
留言列表
