BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:How Watson Learns Superhuman Jeopardy! Strategies - Gerald Tesauro
 \, IBM TJ Watson Research Center
DTSTART:20130927T130000Z
DTEND:20130927T140000Z
UID:TALK47404@talks.cam.ac.uk
CONTACT:Microsoft Research Cambridge Talks Admins
DESCRIPTION:Major advances in Question Answering technology were needed fo
 r Watson to play Jeopardy! at championship level -- the show requires rapi
 d-fire answers to challenging natural language questions\, broad general k
 nowledge\, high precision\, and accurate confidence estimates.  In additio
 n\, Jeopardy! features four types of decision making carrying great strate
 gic importance: (1) selecting the next clue when in control of the board\;
  (2) deciding whether to attempt to buzz in\; (3) wagering on Daily Double
 s\; (4) wagering in Final Jeopardy.  This talk describes how Watson makes 
 the above decisions using innovative quantitative methods that\, in princi
 ple\, maximize Watson's overall winning chances. We first describe our dev
 elopment of faithful simulation models of human contestants and the Jeopar
 dy! game environment.  We then present specific learning/optimization meth
 ods used in each strategy algorithm: these methods span a range of popular
  AI research topics\, including Bayesian inference\, game theory\, Dynamic
  Programming\, Reinforcement Learning\, and real-time "rollouts."  Applica
 tion of these methods yielded superhuman game strategies for Watson that s
 ignificantly enhanced its overall competitive record. \n\nJoint work with 
 David Gondek\, Jon Lenchner\, James Fan and John Prager.\n
LOCATION:Auditorium\, Microsoft Research Ltd\, 21 Station Road\, Cambridge
 \, CB1 2FB
END:VEVENT
END:VCALENDAR