Self-supervised learning

Can a computer learn without labels?

Imagine you're teaching a child to recognize animals by showing them pictures of cats and dogs. You don't tell them what each picture is; instead, you encourage them to guess based on shapes, colors, and patterns.

Self-supervised learning is like that guessing game. The computer tries to figure out patterns in data without being told what to look for, just like the child learning to identify animals.

Example

A computer sees a picture of a cat, then a slightly rotated version of the same cat, and tries to understand what makes them similar.

Remember this

Self-supervised learning allows computers to learn from data without needing labeled examples.

Text adapted from Wikipedia, licensed under CC BY-SA 4.0.

Related concepts

Underdetermination

Evidence may support multiple theories

Ethics of artificial intelligence

Can AI mimic actions but not understand feelings?

Artificial general intelligence

Can we teach machines to think like us?

Private language argument

Private language is unintelligible

Computational theory of mind

Can our thoughts be like computer programs?

Consciousness

Why do we struggle to understand consciousness?

Swipe through 100 ML concepts daily

Open Pocket Polymath