Program · NEU·04
We reverse-engineer the algorithms of the brain, and build the interfaces to it, to make machines that reason more like we do. The brain remains the only system that achieves general intelligence at twenty watts, and we study it as both an existence proof and a blueprint.
What we work on
A program is a small set of hard, coupled problems. These are the ones this group is working on now.
The brain does what our largest models do at a fraction of the energy. That efficiency is not an accident of biology, it is a clue about the algorithms. We study cortical circuits to recover the principles that make learning cheap.
Our interface program builds high-channel-count arrays that record and stimulate cortical tissue with cellular precision. The near-term aim is clinical, restoring movement and memory; the long-term aim is a two-way channel between minds and machines.
Backpropagation is powerful but biologically implausible and hungry. We investigate local, online learning rules, the kind a cortex could actually run, and fold what we learn back into architectures that train more like brains and less like data centres.
Signals from the program
Selected writing from this program
Collaborate
We hire researchers and engineers who want to push one of these programs forward, and partners who want to put the results to work.
