Deep Reinforcement Learning has produced some impressive results—mostly in a particular kind of game or game-like setting—which are worthy of praise. However, we (eventually) want agents which do practical stuff in the real world. Is the real world like these games? What's the realism gap? Are there games that close it a little? All these questions will (possibly) be partially and vaguely answered as I present a new environment for RL research which will help push the boundaries of our field (without setting your cluster on fire), and discuss some recent methods developed to scratch the surface of the challenges associated with it.
Edward Grefenstette was a staff research scientist at DeepMind, following a (short) period as the CTO of Dark Blue Labs, which he also co-founded. Prior to his move to industry, he was working at the University of Oxford's Department of Computer Science, and was a Fulford Junior Research Fellow at Somerville College, while also lecturing at Hertford College to students taking Oxford's new computer science and philosophy course. Before joining Oxford's DCS in 2008 to earn his MSc which led to his DPhil, hedid graduate work in the philosophy departments at the University of St Andrews in mathematical logic, the foundation of mathematics, and some philosophy of language. Prior to this, he obtained a BSc in Physics & Philosophy from the University of Sheffield.
Zoom Participation Enabled. See announcement.