Attention may be all we need, but are Transformers the best Architecture for Attention

Liquid.ai's synaptic modeled NN are much more efficient than standard NN at learning things, and a model closer to biology. The following video by their founder is a great intuitive introduction to how NNs and consequently LLMs work, as well as showing how much more efficient there are architecture is. As an OG in tech, I can tell you that fist out the gate and initial dominance is rarely the long range winner in tech. This work of Liquid.ai's is very interesting.

2 comments