Supervised learning
Supervised learning is a learning method that teaches AI using data with correct answers. This is done by labeling the photo “cat” and training it.
Supervised learning is a learning method that teaches AI by showing large amounts of data matching problems and correct answers. If you solve a workbook that says cat for cat photos and dog for dog photos, and repeat the scoring, the principle is that you will eventually be able to distinguish photos you see for the first time.
This is the most basic and widely used method in machine learning, and is powerful for problems with clear answers, such as classifying spam emails, assisting in disease diagnosis, and predicting demand. Many AI projects in industrial settings are still based on supervised learning, and this method is also used in fine-tuning to refine LLM to fit its purpose.
The biggest burden is that a person must manually label the correct answer. Labeling costs are high, and if the label is incorrect or biased, the AI has the limitation of learning from the error.
✅ Why it matters
- It stably produces high accuracy in problems where the correct answer is clear
- Applies to most AI tasks in industrial settings such as classification and prediction
- It is a basic concept that serves as a reference point for understanding other methods of machine learning.
⚠️ Limits and debates
- It takes a lot of manpower and money to create the correct answer label
- AI learns the errors and biases of the label as it is
- It is weak to new types of data outside the learned range