1.3 What AI Can and Cannot Do
Now that you have met AI and generative AI, the practical question is: what should you give it to do? A useful starting point is a task that is easy to describe, easy to check, and easy to correct if something goes wrong.
AI is often useful for organizing, rewriting, summarizing, explaining, comparing, and creating first drafts. An office worker might turn scattered notes into a notice. A student might request a simpler explanation of an unfamiliar idea. A traveler might begin with a packing list. A person still needs to provide real context and check the result.
Helping is not the same as being responsible
AI did not attend your meeting, does not know a company rule that nobody supplied, and cannot accept responsibility for a wrong contract or unsafe medical advice. It can help prepare material, but responsibility remains with the user and the person making the final decision.
Ask three questions when judging a task:
1. Is there enough information? Has the AI received the context needed for the task?
2. Is the result easy to check? Can you spot a wrong name, date, or conclusion?
3. Is the cost of an error manageable? Is the mistake an easily edited sentence, or could it affect health, money, legal rights, or another person's privacy?
The less information you have, the harder the result is to check, and the greater the cost of an error, the less appropriate it is to let AI decide automatically.
Begin with one low-risk task
Your first attempt does not need to be a complicated project. Choose text without sensitive information and ask AI to do one of these things:
- Make the tone friendlier.
- Turn five scattered points into a list.
- Explain an idea so a child could understand it.
- Draft an editable weekend chore schedule.
Whether you use ChatGPT, Claude, Gemini, Qwen, DeepSeek, Kimi, Zhipu Qingyan, or another product, observe the first response and then add instructions. Products and features change, and this course does not require one particular brand.
The key lesson is not a list of names. It is a sense of boundaries: AI is a tool that can participate in work, not a person who is automatically correct or responsible for you. The next chapter introduces one of the foundations of generative AI—the large language model.