What does it mean? - nothing - beginning of the end - the end What does Jennings say? - echoed by Brad Rutter http://blogs.wsj.com/ideas-market/2011/02/21/why-i-lost-to-watson/ - http://www.slate.com/id/2284721/ ... "an uneasy sense of familiarity" different from Kasparov's sense ... lack of emotion (cockiness, discouragement) and fatigue ... "fields like medical diagnosis, business analytics, and *tech support* are automated by question-answering software" and "the first knowledge-industry workers put out of work by the new generation of 'thinking' machines. 'Quiz show contestant' may be the first job made redundant by Watson, but I'm sure it won't be the last" - http://www.nydailynews.com/opinions/2011/02/17/2011-02-17_ken_jennings_exclusive_oped_jeopardy_champ_says_computer_nemesis_watson_had_unfa.html Claims made by both players - Watson was "stumped many more times than top-level human players" - Watson was actually slower thinking than top-level human players - evidence: short clues killed it Buzzer play seems to be the real key! - speed, precision, and uniformity - and Daily Double strategy follows from dominance on the buzzer - http://ken-jennings.com/blog/?p=2554 - This is a fair advantage: it is what computers are. Human players anticipate instead. - how make a fairer match? : eliminate human turning on the light | make Watson anticipate off voice, too - ... or http://i.imgur.com/lZd0q.jpg [ media/format.jpg ] What does Norvig say? - data trumps knowledge when knowledge is imprecise, contingent and data are plentiful and representative - use rules to smooth off rough edges and fill gaps - "For those who were hoping that a small number of general rules could ex- plain language, it is worth noting that language is inherently complex, with hundreds of thousands of vocabulary words and a vast variety of grammatical constructions. Every day, new words are coined and old usages are modified. This suggests that we can't reduce what we want to say to the free combination of a few abstract primitives." - little worry about overfitting: too much data! and at scale, even rare events have a collective presence -- the long tail of intelligence? - the bigger challenge: pull meaning out of structured data on the web! ----- Meet with teams Friday : midterm deliverable is due Thursday: mostly discussions with teams next Tue: 10-15 minute informal status reports from each team