Overview
Proposed by Alan Turing in 1950, the test involves a human evaluator having a natural language conversation with a human and a machine (without seeing them). If the evaluator cannot reliably tell which is which, the machine is said to have passed the test.
Modern Context
While many modern LLMs can easily fool people in short conversations, the Turing Test is often criticized for focusing on 'imitation' rather than true understanding or consciousness.
Variations
- Total Turing Test: Includes visual and physical interaction.
- Reverse Turing Test: Where a machine tries to tell if a user is a human (e.g., CAPTCHA).