Diana’s Reviews > A Thousand Brains: A New Theory of Intelligence > Status Update

Ok, this discussion of Hebian Learning and how it's wrong is interesting... Mostly because with AI, we do talk about learning through the strengthening or weakening of synapsea (connections between neurons). But this idea that new synapses form and old owns dissappear make me wondering... Could you used this to develop an AI model that is truly deployment aware and learn in real time? Something to think about...

It's interesting to consider Mountcastle's proposal for how the neo cortex is structured and functions given how we've seen learning and emergent intelligence form from traditional neural networks and basic algorithms for machine learning. The ideas of the same units of the neo cortext having completely different functions is exactly what you see with neurons, weighs, and even attention connections in ML and NNs...