Is "learning from humans" the new "layer-wise pre-training"?
I don't know man, our reference intelligence model (us) doesn't do layerwise pretraining, but most of what it knows was learned from humans.
-
-
w8 w8 w8, neural development takes quite some time and is characterized by asinc https://en.wikipedia.org/wiki/Critical_period … across b-areas ~ layerwise pretrain
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
I still think learning from humans is useful & important, but it'll also learn a lot of evolutionary/cultural baggage that it must unlearn.
-
Perhaps machines will learn most of what it knows from other machines more efficiently.
- Show replies
New conversation -
-
-
Aren’t we (humans) actually pre-trained? Our DNA was optimized by evolution and we get that pre-training for free initializing our params
-
In DNA terms, the "source code" for the brain is a few megabytes. That means we contain virtually no hard-coded high-level representations.
- Show replies
New conversation -
-
-
Doesn't it, though? I should think that you did, at some point, study the alphabet, basic shapes, names of colors, types of animals....
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
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.