This was quite interesting. I particularly like the part that started with:
— Jeff Dean (@JeffDean) December 9, 2018
"For example, consider the cost of a single bit in the brain as it travels through a target neuron [18].
...
burning only 2.3 femto-Joules (fJ).
..." https://t.co/UMgAXlpMzz
Good read! Fave quote: “suppose you are playing tennis & hit the ball incorrectly. Which of your 100 trillion synapses are to blame? And how does the brain correct the right synapses in your motor system when the error is delivered through the visual system hundreds of ms after?” https://t.co/SifpnF7rvv
— Andrej Karpathy (@karpathy) December 9, 2018
Great article about how cogsci, psychology, neuroscience played a central role in shaping the AI field.
— hardmaru (@hardmaru) December 10, 2018
Recently, perhaps due to the formalization of subfields in ML (probabilistic generative models, deep RL, etc.), I have noticed a pushback whenever things are “brain inspired”. https://t.co/IXmKt66A6n
