Friday, February 18, 2011

This blogger is impressed... but just mildly... by Watson's streak on JEOPARDY!

So one of the bigger stories this past week has been IBM's supercomputer Watson beating all-time champions Ken Jennings and Brad Rutter during a three-game series on Jeopardy!. Watson's final haul was $77,147, with Jennings coming in second with $24,000 and Rutter third with $21,600. During Final Jeopardy! on the last night of the contest Jennings wrote "Who is Stoker?" and in parenthesis added "I for one welcome our new computer overlords".

I thought Watson's performance was... impressive. But, I'm not too terrified about a droid uprising just yet.

For one thing, Watson is not much more than a glorified search engine. IBM's engineers assured the Jeopardy! audience that Watson could not access the Internet: that "he" relied entirely on his oodles of terabytes of storage, containing (it is thought) every iota of trivia that has possibly been digitized. Clearly, an advantage was held by Watson.

And yet, even that was fallible, as was demonstrated by Watson's widely-scorned inability to know that Toronto is not a U.S. city (the correct question should have been "What is Chicago?"). Watson also reported that Serbia was a country in the European Union (it is not).

Remember when Garry Kasparov beat IBM's Deep Blue in 1997? Kasparov also fought "sequel" Deep Junior to a draw in 2003. Now, to me that is much more extraordinary computer technology, even though those machines never achieved clear victory. There was legitimate strategy and intuitive thinking involved in those contests. In the end, human wetware prevailed over silicon. Watson, as far as I was able to tell, showed none of that capability.

But I will tell you what Watson did have that earned it some respect from this writer: that it was able to, for the vast majority of the time, communicate in natural language as well as most humans.

I first read about the Turing test when I was a high school sophomore. The concept has interested me since: Alan Turing's proposed test for a computer's ability to think. The idea is that if a human can not discern whether he is communicating with another person or with a computer, then that computer has achieved a measure of intelligence comparable to a human being.

What I saw on Jeopardy! this past week, was the most significant demonstration of how close computers have become to passing the Turing test. It's not quite there yet... but it is pretty darn close.

In the meantime, I wonder if IBM could pit Watson against Garry Kasparov in a game of chess? C'mon Garry, the rest of us humans are counting on you to win back our honor! :-P

1 comment:

Anonymous said...

Actually Kasparov lost to Deep Blue in 1997, though he did win the first match in 96.

Friday, February 18, 2011

This blogger is impressed... but just mildly... by Watson's streak on JEOPARDY!

So one of the bigger stories this past week has been IBM's supercomputer Watson beating all-time champions Ken Jennings and Brad Rutter during a three-game series on Jeopardy!. Watson's final haul was $77,147, with Jennings coming in second with $24,000 and Rutter third with $21,600. During Final Jeopardy! on the last night of the contest Jennings wrote "Who is Stoker?" and in parenthesis added "I for one welcome our new computer overlords".

I thought Watson's performance was... impressive. But, I'm not too terrified about a droid uprising just yet.

For one thing, Watson is not much more than a glorified search engine. IBM's engineers assured the Jeopardy! audience that Watson could not access the Internet: that "he" relied entirely on his oodles of terabytes of storage, containing (it is thought) every iota of trivia that has possibly been digitized. Clearly, an advantage was held by Watson.

And yet, even that was fallible, as was demonstrated by Watson's widely-scorned inability to know that Toronto is not a U.S. city (the correct question should have been "What is Chicago?"). Watson also reported that Serbia was a country in the European Union (it is not).

Remember when Garry Kasparov beat IBM's Deep Blue in 1997? Kasparov also fought "sequel" Deep Junior to a draw in 2003. Now, to me that is much more extraordinary computer technology, even though those machines never achieved clear victory. There was legitimate strategy and intuitive thinking involved in those contests. In the end, human wetware prevailed over silicon. Watson, as far as I was able to tell, showed none of that capability.

But I will tell you what Watson did have that earned it some respect from this writer: that it was able to, for the vast majority of the time, communicate in natural language as well as most humans.

I first read about the Turing test when I was a high school sophomore. The concept has interested me since: Alan Turing's proposed test for a computer's ability to think. The idea is that if a human can not discern whether he is communicating with another person or with a computer, then that computer has achieved a measure of intelligence comparable to a human being.

What I saw on Jeopardy! this past week, was the most significant demonstration of how close computers have become to passing the Turing test. It's not quite there yet... but it is pretty darn close.

In the meantime, I wonder if IBM could pit Watson against Garry Kasparov in a game of chess? C'mon Garry, the rest of us humans are counting on you to win back our honor! :-P

1 comment:

Anonymous said...

Actually Kasparov lost to Deep Blue in 1997, though he did win the first match in 96.