SYSTEM NOTICE

Auto translation by AI. Be sure, accuracy, nuances and authorial intent may not be fully reflected.

[AI Terminology Illustrated Series] #5 5 AI Terms to Understand How AI 'Learns' | AI x Information Edition

2026年6月20日 08:09

Hello everyone.Shinya Latte here.

This time, we are covering AI and information.

This time, we will change our perspective a bit and explain 5 terms related to how AI 'learns'—specifically, how it becomes smart in the first place.

This is a must-read for those who don't quite get what 'supervised learning' or 'reinforcement learning' mean when they hear them in the news.

1. Supervised Learning

This is the most fundamental method among AI learning techniques.

By providing a large amount of problems and correct answers (labels) as a set, the AI learns.

For humans, it's like diligently solving a workbook that comes with an answer key. By looking at many examples, the AI becomes able to judge 'this is a cat' even when it sees a new photo.

Many familiar AI applications, such as spam email detection and image recognition, are created using this method.

2. Unsupervised Learning

This is a learning method where no correct answers (labels) are provided at all.

It's like giving the AI only data and asking it to 'group similar things together on your own.' Even without humans teaching it 'this is the correct answer,' the AI discovers hidden structures within the data.

For example, if you provide a large amount of customer data, it will automatically find groups like 'frequent buyers' and 'occasional buyers'.

It is highly effective when dealing with data where the correct answer is unknown or cannot be prepared.

3. Reinforcement Learning

This is a method of learning by repeating 'reward for success, penalty for failure'.

While using trial and error, the AI finds for itself the actions that maximize rewards. The concept is similar to dog training where you give a treat when they succeed.

AI that has defeated professionals in Go and Shogi, as well as robot motion control, are representative examples of this method. It refines its optimal strategy on its own while repeating failures many times.

It takes time, but it can sometimes devise moves that humans would never think of.

4. Zero-Shot Learning

'Shot' refers to the 'number of examples.' Zero-shot refers to having it perform the task immediately without showing any examples.

'I've never seen a zebra, but if it's explained as an animal that looks like a horse but has stripes, I can identify it'—that's the kind of application ability it's similar to.

Asking ChatGPT to 'summarize this text' without showing any examples and having it do so is exactly this. It's a feat only possible because it already possesses a vast amount of knowledge.

5. Few-shot learning

'Few' means 'a small number.' It is a method of teaching it how to do something by showing only a few examples.

While zero-shot is 'no examples,' few-shot is the image of showing '2 or 3 examples.' By showing just a few examples like 'answer in this format,' the AI grasps the pattern and mimics it.

For example, if you provide a few examples like 'Convert as follows: Happy -> 😊, Sad -> 😢,' it will handle the rest in the same format. It is also very practical as a prompt tip.

Summary

✅ Supervised learning
The basic method of learning using data with answers

✅ Unsupervised learning
A method of having it group things on its own without being given the correct answers

✅ Reinforcement learning
A method of improving through repeated rewards and penalties

✅ Zero-shot learning
A method of applying it immediately without any examples

✅ Few-shot learning
A method of having it grasp patterns by showing a few examples

Thank you for reading. See you in the next article.

★
★
★

I have compiled tips on AI image generation, prompt creation, game development, and efficiency improvements.If you think you'd like to see more, please be sure to check out my magazine.

📚 Magazine↓↓
▶ [AI x Information]
I have organized basic AI knowledge, terminology explanations, and how to use generative AI for beginners.

▶ [AI x Illustration]
I have compiled practical notes on AI illustration production, including prompt creation, image generation, composition, color schemes, and character design.

▶ [AI x Parenting/Housework]
A collection of ideas for making life easier by incorporating AI into parenting and housework. I introduce AI utilization methods that can be used at home, such as time-saving techniques and childcare support.

▶ [AI x Chrome Extensions]
A magazine introducing Chrome extensions. It focuses on extensions that improve work efficiency.

▶ [AI x Tools]
I introduce AI tools that can be used for work and daily life, such as ChatGPT, Claude, Gemini, and Chrome extensions.

▶ [AI x Design]
I have compiled ideas for creative production using AI, such as banners, thumbnails, and document design.

Thank you very much.

📱 SNS
X・Threads・InstagramI also post AI illustrations and production notes on these platforms. I would be happy if you could follow me. You can tap each link to go to the SNS.

Late Night Latte | AI x Create
I've been tidying up my note little by little lately, while testing what I can do with AI. Your likes and follows are great motivation for me to keep updating!!

いいなと思ったら応援しよう！

この記事が参加している募集

やってみた

12万件

スキしてみて

84万件