The kicker is: the amount we're investing in science has gone up enormously (think 10-100x) over the same time period, whether you look at $, number of scientists, or number of publicationspic.twitter.com/9gBu2FREw0
Show this thread
-
the algorithmic breakthroughs necessary to get alphazero to the incredible level it got to were extremely conceptual
Alphazero was mostly integrating existing ideas. Very nicely, I might add, and that's not to denigrate it at all (figuring out what to integrate and doing it is very, very hard). But I don't think there was much new there.
-
-
The completly unsupervised reinf. Learning they applied in the last iteration was unprecedented and set a new standard for machine learning
-
TD-Gammon (1992) played at expert level using unsupervised reinforcement learning, and the DeepMind team have of course acknowledged that forthrightly. DM do a nice thing, integrating TD-Gammon's approach with Monte-Carlo Tree Search (also an old idea, which DM acknowledge).
- 1 more reply
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.