IBM Supercomputer v. Humans on Jeopardy! Next Week

Next Mon­day, the long-run­ning Amer­i­can game show, Jeop­ardy!, will air a tour­na­ment of cham­pi­ons, pit­ting its two biggest win­ners, Ken Jen­nings and Brad Rut­ter, against Wat­son, IBM’s newest super­com­put­er. And it will pro­vide an occa­sion to answer an impor­tant ques­tion: Can com­put­ers under­stand the sub­tleties of lan­guage? Can they answer ques­tions when they’re posed in less than straight­for­ward ways? When, for exam­ple, the ques­tions use word­play and puns? IBM worked on the project for four years, and the ear­ly indi­ca­tions sug­gest that com­put­ers can undoubt­ed­ly mas­ter these sub­tleties. (Just watch this Wat­son match against less accom­plished Jeop­ardy! play­ers.) This arti­cle does a good job of explain­ing the fair­ly stag­ger­ing things hap­pen­ing on the back­end of the new IBM com­put­er, and how this research might shape the future of com­put­ing. The Watson/human face­off begins next Mon­day, with two match­es tak­ing place over three days. Once video becomes avail­able, as it inevitably will, we will tweet it on our ever-flow­ing Twit­ter stream.

PS NOVA aired a pro­gram last week ded­i­cat­ed to the Watson/Jeopardy! chal­lenge. If it’s not geoblocked, you can watch it right here. H/T to @eugenephoto for flag­ging the pro­gram.