Biologically inspired language model using Jaccard Surprise as its only training signal. No backprop. No GPU. Online Hebbian learning from corrections. Two-layer cortex with apical feedback. Runs on CPU under 200MB.
python machine-learning ai artificial-intelligence computational-neuroscience artificial-neural-networks neuromorphic htm jaccard hebbian-learning online-learning cortical-learning sparse-distributed-representations cpu-native lightweight-ai biologically-inspired no-backpropagation
-
Updated
May 21, 2026 - Python