Computation with sequences of neuronal assemblies
Talkat Understanding Higher-Level Intelligence from AI, Psychology, and Neuroscience Perspectives · 2024-06-24..06-28·
summary
A concrete claim: a NEMO-style network with random connectivity, k-winners-take-all inhibition and Hebbian plasticity can form assemblies that encode symbols and link into sequences, enabling sequence memorization and online prediction. Simulations build a finite-state-machine recognizer (example: decimal strings divisible by 3) and a Turing-head emulator combined with external tape memory, demonstrating practical computation in the model. Known limitation: no implemented reward signal and plasticity only strengthens.
AI-written summary · about
chapters (9)
- 0:00NEMO and cognitive modeling
- 1:05Neurons, cognition, and algorithms
- 2:29Assemblies in NEMO
- 3:33Sequences of assemblies
- 4:56Finite state machines in NEMO
- 6:44Turing machines for computation
- 7:44Closing remarks
- 8:10Q&A: Biological plausibility
- 10:20Q&A: Reinforcement learning limits
related talks
55mComputation with Assemblies
01h 13mComputation in Networks of Neurons in the Brain II
41mBrain-inspired computational primitives for reliable neuromorphic intelligence
47mCapacity of Neural Networks for Lifelong Learning of Composable Tasks
30mCracking the code: neuronal networks in the brain
43mThe Biological Path Towards Strong AI
51mThe Brain as an Efficient, Robust Adaptive Learner
47mDynamic Neural Network Structures Through Stochastic Rewiring