Essay 2 for 15-103

Aisha Hasan
Essay 2
22/11/2010
The article focuses on artificial intelligence and discusses I.B.M.’s supercomputer
Watson, exploring the field of question answering technologies.
Recently, the supercomputer Watson won mock games of “Jeopardy!” which pitted it
against human players to test its question answering abilities. Watson is unique: unlike search
engines it doesn’t throw up pages of results that may have an answer matching the query, but
rather gives the correct answer itself.
Where previous concerns in the field were insufficient computer memory and speed,
now they are due to the nature of human language (e.g. questions in the game are full of
intended meanings, puns and connotations).
Older “smart” machines have existed – previously I.B.M.’s Deep Blue battled chess
grandmaster Garry Kasparov and won. However, chess is logical and adaptable to maths –
something computers handle well. Ultimately there isn’t a market for machines that “play
chess really well”, but markets for question answering machines do exist in instances where
time is of the essence: law firms, call centres, help desks & hospital emergency rooms.
Watson works by running over a hundred algorithms simultaneously, analyzing a
question in different ways, providing several solutions and taking the solution that turns up
several times to be the most probable. A clear edge Watson has over humans is lack of
emotion: it doesn’t falter under pressure and past blunders don’t affect it.
Limitations also exist: machines like Watson can only provide answers to questions
requiring objective answers (as opposed to judgemental), and since human knowledge isn’t
only textual computers may need to be taught one fact at a time, to be able to use
mathematical reasoning on them. Uncertainty also exists because Watson hasn’t been able to
answer some questions fast enough – however some argue this is because it doesn’t buzz
until it is sure about the answer, whereas humans often buzz first, think later.