this tweet - “real question & ... perception question” (not what the twitterverse is attributing to me) clarifies what think is important here, @egrefen @IntuitMachine & others
-
-
Show this threadThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
There's definitely payoff in Google's business: see for example the datacenter cooling improvements they have achieved with
@DeepMindAI's help. That's a real cash dividend. How many other areas in Google are getting DeepMind's help and squeezing another few % on B$s in revenue? -
that’s discussed in the article i link; the loss is revenue (mainly from that) minus costs
- 1 more reply
New conversation -
-
-
RL does seem remarkably impractical given how much money has been spent on it.
-
Effective application of RL requires tons of data or a good simulator of the system to be managed. Many industries are developing "digital twins" of their systems, and these will enable RL to be applied to those systems.
- 4 more replies
New conversation -
-
-
Yes, General Intelligence clearly has greater human applicability than Grand Unified Theory. One can also say the same about another unsolved problem: Nuclear fusion. Nobody has realistic financial costs to achieve any of these grand challenges.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
I don't disagree with this, but I think we might diverge wrt where to go from there (or maybe not!)
- 1 more reply
New conversation -
-
-
My existing preference is clear in my profile, but I would say RL--the problem--is the right next target. Deep RL is just an approach to RL problems that happens to be the most successful right now. We should be ready to explore other approaches to RL though.
- 3 more replies
New conversation -
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.