No, I’m NOT going to use: “It’s elementary for Watson” as the title of this post! That would just be SO cliched. :(
“So how does it feel to be replaced by GOOGLE?”
This was a question that someone asked me at a party when he found out that I was a Librarian. Not really a question so much as just a blatant put down. I let it go. It was late, I was getting ready to go home. Maybe if I’d had a few drinks, I’d have made an issue of it, but I was driving. The fact that the guy who made that comment was an Iraq War Vet with combat experience might have had something to do with it, too. Besides, it was just a question. Other librarians among you may have realized that it’s probably the most common question/put-down/joke that we get these days. Then we get up on our soap-boxes, say that Google will NEVER replace a librarian, because …
… because peer-reviewed journals, damn it! And Boolean operators, and the reference interview! Furthermore, Human Information Behavior, open-ended questioning, and that one web-site of false information about Martin Luther King that’s really just a front some White Supremacist group. And let’s not forget that anyone can edit Wikipedia! Ha. HA! [fumes]
Really, it comes down the fact that computers simply can’t process information in quite the same as a human being. Oh, they’re QUICK alright. If you play a computer in chess, a game with less “thought” and more “calculation”, and the computer will kick your carbon-based butt all the way back to Azerbaijan. But when you’re dealing with human syntax, and the countless vagaries of human speech; the machines don’t stand a chance. Right?
Last week’s Jeopardy featured a new contestant, one unlike others that had played on the quiz show previously. It was an artificial intelligence program run on a cluster of 90 IBM Power 750 servers and packing over 15 terabytes of RAM. The system was named Watson, after IBM’s founder, Thomas J. Watson. IBM challenged Jeopardy to friendly competition between their machine, Watson, and two of Jeopardy’s most relentless contestants. Jeopardy’s all-time biggest money winner Brad Rutter, and the record holder for most consecutive games won, the “Mormon Assassin” Ken Jennings. These two powerhouses competed against Watson in a three day tournament lasting from February 14-16.
Now, I know what you’re thinking. You’re thinking: “But they humans don’t stand a chance! A computer AI can do billions of calculations per second and potentially has access to the accumulated knowledge of the human species! While the humans are reading the question, Watson will be running through the entire contents of the Encyclopaedia Britannica online!”.
The truth is, it’s not really that simple. As it turns out, human thought is really REALLY complex. The things that you or I might take for granted, are the areas in which artificial intelligence really sweats it out. The clues on Jeopardy are never as simple as they first appear. Sure, intelligence is a factor, but many clues use figures of speech such as puns, idioms, pop culture references, rhyming, and slang. A Jeopardy contestant doesn’t merely need to be “smart”. He or she needs to be able to think on his or her feet. The proper question isn’t so much: “What is the correct response”, as: “Where are they going with this clue?”. A recent PBS special on the series: Nova, explained some of the difficulties in teaching artificial intelligence to mimic human thought. A simple word like “milk”, is something that children learn before they’re even old enough to go to Kindergarten. But the very concept of the word: “milk” has an elaborate framework connected to it that the human brain is specialized to understand. We humans know milk to be a white liquid produced in the bodies of mammals to feed their young. We also know that it’s usually a white liquid, that humans take the milk of other mammals to be packaged and sold in supermarkets. We know how it tastes, when it’s gone bad, and that one of the greatest experiences of human existence is a glass of milk served icy cold with a stack of fresh baked chocolate chip cookies. We know this instinctively when we hear the word “milk”. A computer has to search for this information. The PBS special describes how Watson got one question wrong, through a simple, almost “human” mistake. The clue was: “In 1698 This comet discoverer took a ship called the paramour pink on the first purely scientific voyage“. The correct response was: “Edmund Halley”. (really, how many comet-discoverers do YOU know?) But Watson picked up a description of the 1975 film: “The Return of the Pink Panther” which described one character in the film as a “paramour”. So between the copious mentions of the words “pink” and “paramour”, Watson decided that the comet discoverer who embarked upon the first purely scientific sea voyage was: “Peter Sellers“. Now, I’ve never made it onto Jeopardy myself, (despite numerous attempts at the online Jeopardy test) but I can say for certain that Inspector Clouseau never discovered a comet.
Warning! Spoilers Below!
Watson was put through a rigorous series of mock Jeopardy challenges, first against the programmers at IBM, then against a series of former Jeopardy contestants. This wasn’t merely a test to see if Watson could hold its own on Jeopardy. This was a learning experience for Watson. Watson’s sophisticated computer brain allows it to actually learn from its mistakes.
As Watson continues to fail against its human opponents, it learns more and more about the complexities of human thought and speech. Once inside the competition, Watson was unable to access the Internet to look up answers. Watson would be forced to rely on its pre-existing knowledge base, just like a human. The longer Watson practiced, the smarter it got.
In the end, the oversized pile of silicon and metal managed to beat its pathetic human competitors. It beat them handily. Every so often, Ken and Brad would start to rally, and things started to look better for the human race. Each response provided an insight into the mind of Watson. The questions that it got wrong were wrong in such a way that it was logical to see exactly why Watson got those particular clues incorrect. But by the final round, Watson was far ahead of the human contestants and could not be caught. Jennings and Rutter took it in stride, but the results were imposing.
So what does this mean for us poor squishy humans? According to the IBM team who created Watson, this competition was still a win for the humans. Watson was not conceived in a vacuum. The artifical intelligence that is IBM’s Watson is nothing less than the result of 7 years research and development by brilliant humans working in IBM’s research division. The mere fact that humans can actually design and build a machine, a tool, out of metal and wires and silicon that comes this close to human thought, could be considered a resounding victory for humankind. How much longer until we actually have machines to speak to? Machines that can empathize with human emotions? How much longer until we have machines that do all of our thinking for us, as we sip margaritas on a beach in Cabo trying to think of the next big step in human advancement?
Will these machines ever replace librarians? Well at this rate, that could be a very real danger. Will there come a day when your friendly local librarian will be replaced by a metal box and a microphone for you to read queries into? Let’s just say that I wouldn’t bet the ranch on that ever happening.
They’ll probably have all risen up against us by then. [nods]