Topic
All digests tagged Neural Networks
Suggested route
Move from earlier context through the available material to the newest signals. This route is a reading aid, not a difficulty ranking.
The lecture explores how computational models, from simple LN models to CNNs, can be used to achieve mechanistic interpretability in the retina, revealing how complex biological functions are encoded.
Read article
Liquid AI is pioneering the next generation of AI by developing Liquid Neural Networks (LNNs), inspired by biological systems. They use meta AI to search for hybrid, hardware-efficient architectures for reliable on-device intelligence.
Read article
This lecture explores Automatic Differentiation, proving that the gradient computation (backward pass) in neural networks is efficient. By viewing networks as differentiable circuits, backpropagation can be computed in time proportional to the number of parameters.
Read article
Interpretability is the neuroscience of AI, aiming to reverse-engineer how neural networks learn from data. Techniques like Probing and Sparse Autoencoders allow us to map meaning onto model activations, which is essential for building safe AGI.
Read article
We found a set of representations in Claude’s neural activity that play a similar role to conscious thought. We call this collection the J-space.
Read article