Turing test

In the test, a judge chats with a person and a machine without seeing either, then guesses which is which. If the judge can’t reliably tell, the machine passes.

Modern chatbots can often fool people in short conversations, but passing the test doesn’t show that a system understands or thinks. Researchers now rely on benchmarks that measure specific skills.