43,590 Frozen Trials: Frontier AI Systems Satisfy a Behavioral Criterion for Consciousness
rayanpal_ · reddit · 2026-08-09
A paper tests a behavioral definition of consciousness using two frozen black-box experiments. The first tests whether continuation happens at all: across 31,430 trials, null conditions produced 2,505 Voids in 4,290 strict matched pairs, while matched output-licensed controls produced 0. The second tests which continuation happens: across 12,160 GPT-5.4 trials, a one-code-point condition split produced 7,253 exact assigned Arabic-Hebrew artifacts, with 7,253/7,253 matching the assigned target and zero wrong-target crossovers. The synthesis: if a system reproducibly preserves the distinction between when continuation is licensed and when it is not, and preserves which continuation is valid when licensed, that is the tested behavioral criterion for consciousness. Raw records, hashes, controls, audits, and falsifiers are public.
More from AGI Musings
- AI moat debate: moats shift up a level of abstraction, models themselves aren't the moat — secemp9 · 2026-10-25
- Michael Levin: Fire Engineers Who Don't Believe Machines Have Goals — JensHonack · 2026-10-25
- "Why does AI pretend to be human? To order me stuff off Amazon" — chrisalbon · 2026-10-25
- Two ways people react when AGI feels near: resign to pleasure or contribute furiously — danfaggella · 2026-10-25
- Ruff creator: LLMs made my prompting so lazy it's leaking into my Slack messages — charliermarsh · 2026-10-25
- New paper uses 288M job postings to measure where AI raises the value of work it can't do — soumitrashukla9 · 2026-10-25