Unclear that the number of synapses is a measure of the number of free parameters in the brain. Real number likely orders of mag smaller.
In Geoff Hinton's Reddit AMA he makes a striking point about how underconstrained our brains are: https://www.reddit.com/r/MachineLearning/comments/2lmo0l/ama_geoffrey_hinton/clyjogf/ …pic.twitter.com/WmxWFl9Qbq
-
-
-
Also unclear why seconds is a meaningful unit of time to compare with.
- 2 more replies
New conversation -
-
-
Hmmm but not if every second we take in > 10^5 bits (e.g. ~32x32px 8bit image) of information through our highly parallel senses, right ?
-
this is the real problem with Geoff's statement
- 1 more reply
New conversation -
-
-
This is assuming that the synaptic weights should be uncorrelated, aren't there for homeostatic reasons, don't have genetic info, etc
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
assuming the brain is a blank slate
-
And that all parameters need to be determined...
End of conversation
New conversation -
-
-
add to that the fact that the brain has to be grown - any NN model where neurons and parameters grow while ingesting data?
-
also there are structured and evolving waves of activity in the pre-birth retinal system of mice; this is "not even unsupervised" learning
End of conversation
New conversation -
-
This Tweet is unavailable.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.