Technology has grown adept at
playing games. Everyone's mobile device
can play solitaire, you have to be skilled to beat computer chess even on
"easy", and Olin students are writing Sudoku solvers and Rock-Paper-Scissors
bots that are impossible to beat. IBM
took this a step further when they created Watson, the Jeopardy! playing
computer. Jeopardy is a quiz game where
three contestants choose a category, and a monetary value, and are given a
"clue", which is the answer to a specific question.
At first glance, this doesn't seem
a truly impressive undertaking; computers can store and retrieve data better
than humans, so of course it could just make a cross reference and look up the
correct response. However, things we
take for granted as humans, like "looking up" or "cross referencing" take a
huge amount of programming and data analysis capability for a computer. Even more complicated than computation, data
finding, or character matching, is the heuristics involved in things like
teaching a computer that "Dave" and "David" are statistically synonymous. After years of hard work, however, IBM created
a computer that they thought might beat the Grand Champions of Jeopardy!
On February 14, 2011, Watson went head to head with the two most famous
Jeopardy! veterans: Ken Jennings and Brad Rutter. Watson won. By a landslide. But what now?
IBM has made huge advances in data analysis with Watson, but what can it
do besides annoy Alex Trebek? It was
this question that brought PGP Director Sally Phelps, myself, and about a
hundred other academics to the Watson Research Center in Yorktown, New York
last Thursday to attend IBM's Transforming
the Industry: Watson in Education conference.
Click here for part 2.