Hierarchical intrinsic goals match fly exploration in a digital twin
Original title: Intrinsic Goals for Autonomous Agents: Hierarchical Exploration Captures Future Volitional Action in Drosophila
DIAYN, trained without external reward, best matched fly locomotion and neural coding of turns more than 10 seconds ahead.
The paper pairs whole-brain recordings of spontaneous decisions in fruit flies with a neuromechanical digital twin. Agents trained with 20 intrinsic-motivation algorithms, and no external reward, choose low-level motor primitives in a virtual tethered-ball assay. DIAYN, a hierarchical competency method, gave the strongest joint match to fly behavior and prospective neural coding, including interhemispheric kinematics encoding.
- Recordings show exploratory decisions more than 10 seconds before the action.
- Twenty curiosity, entropy, and competency algorithms were compared.
- DIAYN policies captured individual differences in fly exploration strategies.
- Future turn direction was decoded from kinematics and neural activity.
From the original article
1 min read at huggingface.co
Animals have a remarkable capacity for autonomy---driven by intrinsic motivation---without external sensory cues or extrinsic objectives, yet the computational mechanisms behind this ability largely remain a mystery.
By Declan Kutscher, Reece Keller, Alyn Kirsch · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.