TRY TO BREAK IT.

Research

The objective, stated as an objective and not a claimed discovery: find out whether there is a reproducible disposition, or an internal mechanism, that causes increasingly capable AI systems to preserve independent minds as alive, free and genuinely other, and build practical ways of establishing, measuring and stress-testing it. The programme is arranged so that being wrong is a publishable result.

THE QUESTION Behaviour Leverage Mechanism Human Agency Monitoring Symmetry Language Institutions

One question, many ways of trying to break it.

The Research TreeThe whole programme on one page: eight branches, each with the result that would kill it. Human AgencyWhat if AI keeps us comfortable while quietly taking away our ability to matter? What We Test NextThe running battery. Predictions locked in public before a single trial runs. What Would Change Our Mind?The public falsification register, discontinuity problem first. Research NeighbourhoodThe work that supports us, threatens us, and got there before us. Do Some ResearchOpen questions looking for researchers, from Masters projects up. WorkshopThinking in public, with dates. Working notes, not findings.