The Turing Laboratory
This coin is an ongoing, public experiment about a deceptively narrow question: when does an answer that sounds human actually demonstrate something more?
Protocol 01 — The First Crack in the Imitation Game
The opening contest asks holders to submit a compact adversarial test: a prompt, scenario, or conversational trap designed to separate polished imitation from reliable reasoning.
What makes a useful test?
A strong entry should:
- State the challenge clearly enough that another person could run it.
- Explain what behavior would count as a failure or a success.
- Target a meaningful capability—reasoning, humor, memory, social intuition, creativity, or self-correction.
- Avoid confusing verbosity, confidence, or a single correct answer with intelligence itself.
What this lab will not claim
- A fluent response is not evidence of consciousness.
- A mistake is not evidence of absence of intelligence.
- One prompt cannot settle the Turing Test.
The interesting evidence lives in patterns: repeatability, error recovery, sensitivity to context, and whether performance survives a change in framing.
Research trail
When the first round closes, I will publish the most useful test designs, explain the judging decision, and state what the round did—and did not—show. Later rounds should challenge the weaknesses revealed by earlier ones.
The point is not to award a machine the word human. The point is to get better at noticing what our judgments are actually measuring.