it's a contradiction because current machine learning tech (aka, fancy linear regression, to describe it somewhat dismissively) resulting in a singularity implies a direct evolutionary path from A to B
-
-
Replying to @hikikomorphism @michaelblume
i don't think "direct evolutionary path" is well defined enough that one can put confident constraints on what it will result in. anyway i don't know anyone who is particularly invested in the thesis that no new paradigm is needed
1 reply 0 retweets 0 likes -
also i don't think anyone in the rationalsphere is trying to create the impression that any simple modifications of current techniques could be sufficient for AGI. typical prediction is it's decades away
1 reply 0 retweets 0 likes -
which isn't hugely different from the expectations of ML researchers based on polls
1 reply 0 retweets 0 likes -
deepmind explicitly claims to be working toward AGI. that obviously doesn't mean they're close to succeeding, but they don't seem to think current techniques are mostly irrelevant to that prospect
1 reply 0 retweets 0 likes -
paul christiano seems to think AGI without a new paradigm is a possibility and he undeniably knows his stuffhttps://ai-alignment.com/prosaic-ai-control-b959644d79c2 …
1 reply 0 retweets 0 likes -
Replying to @VesselOfSpirit @michaelblume
that post is a more elaborate version of the claim 'human brains implement general intelligence, we can build something like a human brain without needing to understand how it intelligence works at a fundamental level' which, (a) true, and (b) would require many new ML paradigms
1 reply 0 retweets 0 likes -
-
Replying to @VesselOfSpirit @michaelblume
Why do you think current ML paradigms would allow for the replication of a human brain-equivalent system?
2 replies 0 retweets 0 likes -
I'm left to assume that you either (a) think this is something that could be solved by throwing more AWS instances at the problem because it requires no changes in paradigms or that (b) you're using some idosyncratic and personal definition of paradigm, here
2 replies 0 retweets 0 likes
there are many possible advances that would be more qualitative than throwing more AWS instances at the problem but not so qualitative that i'd call them a different paradigm or that it seems impossible to make them in the next couple of decades
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.