“Beyond inspiration: Five lessons from biology on building intelligent machines”
Since he’s cited in some HTM papers, and parts of his talk overlap with HTM theory I thought people here might find it as interesting as I did. He also has some ideas that diverge a bit from HTM.
PDF with his slides is also on that page.
Also worth a watch is Surya Ganguli’s talk: “Theoretical neuroscience and deep learning theory”
A quote I liked: “LSTMs are a step backwards in my perspective from the perspective of modelling neuroscience… Individual neurons don’t have forget gates and write gate and read gates, right, so if we want to compare the dynamics of an LSTM, and understand how function emerges from individual neurons that are not that smart, LSTMs can’t answer that question… Bruno [Olshausen] is vigorously nodding his head.”