I totally agree! Personally, I think deep meta-learning offers a path forward for on this front, though I'll admit the current SOTA for these techniques is quite unimpressive. Thanks for the concrete suggestion :)
progress towards AGI would be reflected in making progress on open-ended domains, eg transferring between video games with different skins but common logic. alas the fact that people are working hard doesn’t mean they progress is being made (cf alchemy).https://twitter.com/zergylord/status/997915182075662336 …
-
-
-
As an example, Matching Networks can perform one-shot classification of dog breeds. Classified lots of other things before, but never dogs. It hasn't extracted the abstract concept of classification, but it feels like progress. https://arxiv.org/abs/1606.04080
End of conversation
New conversation -
-
-
There are plenty of examples of successful transfer learning that seem analogous to yours but more useful - c.f. Google's "interlingua" demonstrating that learning in unrelated pairs can improve each others' translation: https://arxiv.org/abs/1611.04558
-
I think these notions of narrow / general systems are poorly defined. Without an attractive and operational definition of progress towards AGI, you're going to get stuck trying to play arbiter about whether individual lines of research are promising or not.
End of conversation
New conversation -
-
-
> transferring between video games with different skins but common logic We can't do this now? That's surprising to me! (as a non-practicing observer of AI/ML research)
-
We indeed can't do this automatically. To clarify, one system could learn either, but having learned one doesn't help it at the other. What led you to think we could?
- 3 more replies
New conversation -
-
-
Is the shorthand of "Problem solving in open ended domains" AGI? Or human level intelligence?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
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.