Untangled: AI, explained from scratch

There is no agreed test for understanding.

Behavior gives us evidence, not direct access to meaning or experience. Following rules might be all a system does, or understanding might belong to the complete system rather than any one part.

A sharper question names what is being tested. Is it correct behaviour, meaning, adaptability, or conscious experience?

The next module returns to history. Early AI had real successes, but many problems refused to fit inside neat rules and searchable paths.

The whole chapter, simply

A machine can give the right answer without telling us whether it understands.

We can test what it does, but we still cannot see what, if anything, it understands or feels.

The question is still open.

There is no agreed test for understanding.

Behavior gives us evidence, not direct access to meaning or experience. Following rules might be all a system does, or understanding might belong to the complete system rather than any one part.

A sharper question names what is being tested. Is it correct behaviour, meaning, adaptability, or conscious experience?

The next module returns to history. Early AI had real successes, but many problems refused to fit inside neat rules and searchable paths.

The whole chapter, simply

A machine can give the right answer without telling us whether it understands.

We can test what it does, but we still cannot see what, if anything, it understands or feels.

The question is still open.