Unlike less fortunate scientific fields that have unresolvable century-long debates, ML has an easy debate-resolving procedure: just build a system that's obviously more intelligent than its predecessors. Whatever ideas used to build it are the right ones.
blurring distinction between evolution & learning trivializes both, @ilyasut, as I discussed in https://arxiv.org/abs/1801.05667 . Key point: "Evolution .. is a means towards building machinery with embedded prior knowledge, not an alternative to prior knowledge" cc @primalpolypic.twitter.com/iOJ4UsGUKI
-
-
-
totally agree
@ilyasut, but caution that nobody has right system now & best current systems may not be on right track; short term winners may not win out in long-term. road to AI littered w abandoned fads with no staying power, partly because value of short term fruit overrated.
End of conversation
New conversation -
-
-
Formalize both and the distinction. „Wovon man nicht sprechen kann, darüber muss man schweigen."
-
my favorite Wittgenstein quote. possibly relevant for the study of consciousness?
- 1 more reply
New conversation -
-
-
Why not just say that learning is "change"? That pretty much gathers up everything and we can call this a solved problem.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Yep. Will review the paper, but I already agree.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Whatever innateness is, a 'post-innate' layer is a useful construct, before what we see as social/intellectual learning– what a fetus and infant do- e.g. TOM/type-token diff/continuity/causality get built onto innateness to initialize a std learning schema. Firmware, if you will.
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.