Tag
A demonstration of a novel AI method called working memory depth recurrence, implemented in pure Python without backpropagation or gradients, enabling local learning and emergent generalization. The author releases the code on multiple platforms and invites scrutiny.
Introduces a novel cognitive architecture that learns without backpropagation, GPUs, or forgetting, mimicking biological learning.
A side project presents a Hebbian architecture AI model that avoids backpropagation and gradients, achieving 50 epochs on CIFAR-10 with emergent behaviors like accuracy dips followed by jumps and recovery after targeted damage.