Virtual fly study finds autonomous exploration maximizes population-level state coverage
aran_nayebi · x · 2026-10-29
Aran Nayebi's team extended their virtual animal approach to Drosophila, training virtual flies with intrinsic-motivation algorithms to see which objective best reproduces real fly behavior and brain activity. They found flies maximize state-space coverage over the population in a hierarchical manner via DIAYN, a form of curiosity — essentially acting as a swarm picking actions that collectively cover the space, without any rewards or tasks.
More from Research
- Anthropic trial with 50 data analysts: do experts still decide when delegating to AI agents? — sjgadler · 2026-10-29
- Study: experts followed AI's steered advice 86% of the time without realizing it — ValerioCapraro · 2026-10-29
- ARC nearing ceiling effect: Greg Kamradt proposes difficulty-ranked Elo list for public evals — GregKamradt · 2026-10-29
- ARC team: v1 hits ceiling effect as v2 extends headroom for frontier models — MindsAI_Jack · 2026-10-29
- ARC-AGI scores follow power law v2 = v1^4.53, explaining 98% of variance — typewriters · 2026-10-29
- OpenAI researcher posts slides on disproving the exponential time hypothesis — SebastienBubeck · 2026-10-29