This Brain Discovery Could Unlock AI’s Ability to See the Future
1 min read
Summary
Researchers at the Champalimaud Foundation, Harvard University and other institutions have discovered a previously hidden universe of dopamine signalling in the brain.
Single dopamine neurons were recorded as mice learned a new task, and it was found that they don’t just track rewards, but also keep tabs on when they come and how big they are.
This builds a mental map of near-term and far-future reward possibilities, mirroring some emerging AI systems that rely on reinforcement learning.
In several simulations, AI equipped with this multidimensional map better handled uncertainty and risk.
The results could be used to design more efficient reinforcement learning AI that can better adapt to uncertainty, and to treat impulsivity in neurological disorders such as Parkinson’s disease.
