AI Made Friendly HERE

A clever way to leave even prompt engineering to AI.

To use ChatGPT effectively, you need to learn prompt engineering.

You sometimes see people say things like that.

Prompt engineering is the technique of refining instructions to elicit the expected response from an AI.

Certainly, giving appropriate instructions makes it easier to get the desired results.

However, there is one question here.

Couldn’t we just leave that prompt engineering to ChatGPT itself?

ChatGPT is evolving

ChatGPT evolves every time a new model is released.

As its ability to understand instructions and its reasoning capabilities improve, the way to write effective prompts also changes.

Processes that previously required specifying detailed steps might be achievable with new models just by stating the goal.

In other words, prompts that were effective for older models are not necessarily optimal for new ones.

If that’s the case, do we need to relearn how to write prompts every time a new model comes out?

That sounds quite tedious.

Asking ChatGPT how to write prompts

So, let’s change our perspective.

If you don’t know how to give instructions to ChatGPT, you should just ask ChatGPT itself.

For example, you can input the following.

Windowsで動作する画像処理アプリを作りたいです。

あなたに開発を依頼するには、
どのような指示を出すと効果的ですか?

必要な情報があれば、先に質問してください。

That’s all.

Have ChatGPT ask you for the necessary information, and as you answer, you can flesh out the details of your request.

If there are too many questions, you can add, “Please ask one question at a time.”

If you don’t know how to achieve something in the first place, you can start by consulting with the AI, saying, “I want to do this, is there a good way to do it?” instead of asking it to create a program from the start.

Have it think about completion conditions too

When requesting work, completion conditions are also important.

For example, with an image processing app, you need to consider not only that it can process normally, but also how to handle cases where the file is corrupted or saving fails.

However, users may not always think of all the conditions from the beginning.

Therefore, make your request as follows.

このアプリの完成条件を考えてください。

実際に使用することを想定して、
確認するべき項目を挙げてください。

私が決める必要のある項目があれば、
質問してください。

By doing this, you may be able to notice conditions you had overlooked.

Finally, if you have it summarize the consultation content, the prompt for requesting the work will be complete.

Do you not need to learn prompt engineering?

Of course, ChatGPT cannot always suggest the optimal prompt.

For advanced processing or tasks where stable results are required, human verification and adjustment are also necessary.

However, if you use ChatGPT on a daily basis, you may not need to learn prompt techniques in detail from the start.

Explain what you want to do in plain language.

Have ChatGPT ask you for the necessary information.

Finally, have it organize the information into an effective prompt.

Instead of humans mastering prompt engineering, leave prompt engineering itself to the AI.

As ChatGPT evolves, this method is expected to become even more convenient.

Have the AI think about how to use the AI effectively.

Isn’t this also a clever way to do things?

Originally Appeared Here

You May Also Like

About the Author:

Early Bird