Reverse-Engineering the Wetware: Spiking Networks, TD Errors, and the End of Matrix Math
Summary
Pedro Teixeira argues that biological neural processing relies on predictive coding and local learning rules rather than global backpropagation. He explains STDP, dopamine-driven TD errors, and the hardware implications, making the case for neuromorphic chips and brain-inspired AI approaches. The piece also points to practical tools like Nengo and pymdp for experimenting with spiking networks and predictive coding.