I enjoyed your piece, but what are your thoughts on an almost "return to analog" in neuromorphic chips/computing designs or archs. ?
How would we ever have AI be able to run a similar "visual imagination" engine as we do , which helps our reasoning/ideation/world model greatly?
On current architecture, this is so difficult and expensive. Would it make sense to switch our entire computing base from digital? (in certain crucial settings, rather than trying to widely "rewrite everything" only -- and end up remaining too under specialised in hardware.)
Super-interesting topic, but far-enough from my expertise that I don't have anything to say about it! My own push is toward *less*-neuromorphic architectures based more on formal logic, which may indeed be a bad fit for vision.
I am aware of cool stuff going on now in analog computing in the research group of Stanford Prof. Sara Achour.
I enjoyed your piece, but what are your thoughts on an almost "return to analog" in neuromorphic chips/computing designs or archs. ?
How would we ever have AI be able to run a similar "visual imagination" engine as we do , which helps our reasoning/ideation/world model greatly?
On current architecture, this is so difficult and expensive. Would it make sense to switch our entire computing base from digital? (in certain crucial settings, rather than trying to widely "rewrite everything" only -- and end up remaining too under specialised in hardware.)
Super-interesting topic, but far-enough from my expertise that I don't have anything to say about it! My own push is toward *less*-neuromorphic architectures based more on formal logic, which may indeed be a bad fit for vision.
I am aware of cool stuff going on now in analog computing in the research group of Stanford Prof. Sara Achour.