On the alleged impossibility of innate symbol-manipulation, these two papers suggest otherwise http://www.pnas.org/content/105/37/14222 … and http://science.sciencemag.org/content/359/6381/1263 … Meanwhile I would suggest that arguing purely on the basis of tweet, blog posts, and hearsay < idealhttps://twitter.com/tyrell_turing/status/1003315128584925184 …
-
For *engineering* new forms of intelligence, we might want to distill that evolutionary knowledge and write it into our programs directly -- without modeling the "learning" process of how it got there in the first place. That seems to be the debate here.
indeed. why start from scratch recapitulating the first billion plus years of evolution?
-
-
Definitely not from scratch! But it depends on how well we think we've understood that evolutionary knowledge, the component innate to individuals. Trying to understand better always involves some generate-and-test (evolution-like) exploration, which we might want to automate.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
I agree: we don’t wanna start from scratch. But, maybe good NN architectures are the best way to build some good priors in, IMO.
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.