Paradoxically, DeepMind is in many ways a very old-fashioned AI lab, for example doing most of their research in games and simulated worlds. It's an open question whether, once exploiting backprop and GPUs has run its course, they will run into the same wall those labs did.
-
Question is if reinforcement learning is indeed the Master Algorithm as per your book!
-
-
What approach are you more convinced of?
-
My bet is something like https://worldmodels.github.io/ where the point isn't the RL, but everything that happens before. RL is relegated to just a theoretical problem formulation, and you optionally slap an RL module after your predictive model proves to be good
- Näytä vastaukset
Uusi keskustelu -
Lataaminen näyttää kestävän hetken.
Twitter saattaa olla ruuhkautunut tai ongelma on muuten hetkellinen. Yritä uudelleen tai käy Twitterin tilasivulla saadaksesi lisätietoja.