The Jeopardy! winning machine creates only the illusion of intelligence, writes Ian Kerr. But maybe that's the point
Ottawa Citizen, 19 February 2011.
Alex Trebek: Astoundingly, you’ve hit the Daily Double once again! Here is your clue.
Four years in the making, this romantic event witnessed a two billion dollar cluster of Power 750 servers, operating at more than 80 Teraflops, transform two human game show champions into a couple of robots.
What says our reigning champion?
Watson: What is IBM’s St. Valentine’s Day massacre?
So what’s the big deal? Well, for one thing, until Wednesday night, use of the word “human” in Alex Trebek’s clue would have been redundant.
Or, as the former all-time champion, Ken Jennings, put it in his live blog for the Washington Post: “[A]re you kidding? I AM PLAY-ING A PRIME-TIME GAME SHOW AGAINST A SUPER-AD-VANCED ROBOT! This is the coolest thing I will ever do in my life by a factor of a million. The future is here.”
And what exactly did the smartest-ever human competitor mean in his Final Jeopardy! response when he said: “I, for one, welcome our new computer overlords”?
I leave that question to the futurists.
One of my favourite living futurists is Ray Kurzweil. Back in the mid-’80s he predicted that a computer would defeat the world chess champion by 1998. The rumblings began in 1996. Although world champion Gary Kasparov won the match that year, IBM’s Deep Blue did manage to win a game for the first time ever. “I could feel -I could smell -a new kind of intelligence across the table,” Kasparov wrote in Time magazine after the match. But, he went on to say: “Although I think I did see some signs of intelligence, it’s a weird kind, an inefficient, inflexible kind that makes me think I have a few years left.”
Good thing Kasparov did not make his living as a futurist. The next year, an upgraded Deep Blue defeated the grandmaster.
It is perhaps no coincidence that Ray Kurzweil -the man who predicted Kasparov’s demise right down to the very year it would happen -owns this week’s cover of Time for his outrageous prediction that 2045 is the year we will become immortal: the year when humans merge with the machines.
Though hindsight is 20-20, it seems to me that Kurzweil’s 1980s prognostications were much wiser and far less dramatic. Among other things, he forecasted that the IBM victory would generate the following permutation: “we would either think better of computer intelligence, worse of human thinking, or worse of chess.” Using history as his guide, he predicted the latter of the three as the most likely. And, as history has since revealed, he was not only right about that, but likely right about the other two as well.
So much the worse for the future of Jeopardy! as the threshold for human intelligence.
But I think people have totally missed the substantial social significance of what is going on with artificial intelligence projects such as Deep Blue and Watson.
It’s not about whether machines are better than humans at chess or Jeopardy! (or poker, or natural language or any other activity that can be said to involve human thinking). When an amazing company like IBM invests two-ish billion dollars in order to win a measly million-dollar prize, clearly something else is at stake.
Philosopher-mathematician Alan Turing is clearly the progenitor of all of this. In his classic 1950 attempt to answer the question “Can machines think?” Turing created a test that almost all subsequent generations of scientists have striven to pass.
The “Imitation Game,” as he styled it, imagines a competition involving a man, a woman and an interrogator. Separated from the other two, the interrogator asks a series of questions using a textbased device to determine which of the other two is the man and which is the woman. In answering the interrogator’s questions, both contestants attempt to obfuscate their gender. Turing then imagines that we substitute a machine for either of the two players. The object of this new game is to see whether the machine is capable of conversing in a manner that is indistinguishable from its human interlocutors. If it could fool the interrogator roughly as often as the humans could, says Turing, one can claim that the machine can think.
As many artificial intelligence aficionados have already made clear in the blogosphere, both before and after the match, Watson’s victory does not meet the threshold of Turing’s test. Not even close. Watson proved itself as an answer-generating machine par excellence. And, although Watson showed considerable computational skill in its ability to parse natural language, a conversationalist Watson was not.
This reality was not lost on the IBM team. Hence the corporate choice to play Jeopardy! rather than address Turing’s challenge head-on.
So, if IBM’s grand challenge was not to pass Turing’s test, what exactly is going on here?
Operating 16,000 times faster than my laptop, Watson generates responses to clues quickly, autonomously and, as IBM programmers came to learn, unpredictably. This incredible accomplishment in the field of artificial intelligence was practically unimaginable just a few years ago when Deep Blue beat Kasparov at chess.
But Watson can’t really play Jeopardy! -not without a human puppeteer pulling strings behind the scenes. Even if we say that Watson knows how to talk (it’s a stretch), Watson doesn’t know when to talk. An operator is placed offstage, playing the crucial role of sending commands that prompt Watson when to speak, when to answer, when to choose a category or clue, and when to place a bet. It is the human puppeteer who, with our imaginative co-operation, creates the illusion that Watson is playing a game with humans. Without the subterfuge of human intervention, Watson remains a computational instrument -not a Jeopardy! contestant.
Again, I say this not to denigrate IBM’s achievements but to lay the foundations for understanding how IBM’s end game relies on creating the illusion.
Viewers who paid attention to the documentary clips interspersed throughout the threeday exhibition may have noticed IBM’s development team toting shiny silver Macbooks and not the black bento box “ThinkPads” that IBM had become famous for. Recognizing the limited shelf-life of today’s cumbersome personal computers, Big Blue sold its ThinkPad to a Chinese manufacturer in 2005. The Mac presence is, to me, symbolic of IBM’s abandonment of little black boxes in favour of its original namesake: international business machines. The kind of super-machines that will run all of our next generation devices from the clouds.
Maybe it was no blunder after all when the namesake of the newly minted Jeopardy! champion -IBM President Thomas John Watson -allegedly said in 1943 that, “I think there is a world market for maybe five computers.” Once derided as a comment lacking foresight, it finds a new air of prescience on the heels of the mechanical Watson’s Jeopardy! victory.
IBM recognizes that society’s investment in super-machines -be it governments and citizens, health-care providers and patients or the financial sector and consumers -will require various levels of trust. It is not surprising, then, that Watson’s team employed state-of-the-art techniques in the field of affective computing (the goal of which is to synthesize emotion in machines and, at the same time, elicit emotional reactions in humans) to build a human connection between Watson and its audience. Watson’s avatar and voice were endearing and well chosen. It was quite purposefully gendered as male -but not the threatening variety. (Female voice-bots only answer phones, right Bell?) Although Watson’s behaviour and attributes are illusory, the trust they engender is real -further reinforcing Turing’s account.
A quick romp through the sphere-o’-twits reveals what human-computer interaction specialists and the folks in Hollywood have understood for years: audiences can be engineered to like, love, lust for and trust machines like Watson better than human counterparts.
And this is exactly what companies like IBM are counting on. Here is what it said on the front page of its website this week:
“The challenge is over. Watson, Ken Jennings and Brad Rutter concluded their final round of Jeopardy! and the winner was … resoundingly, humankind. Watson’s advances in deep analytics and its ability to process unstructured data and interpret natural language will now be applied to humanity’s most vexing problems. If we can teach a computer to compete on Jeopardy! what could it mean for science, finance, health care and the future of society?”
Here, we have the Californication of computing.
When asked in the Year 2045, some distant, immortal relative of the Watsons is sure to remember the lyrics to that ol’ Chili Peppers’ song:
“Space may be the final frontier but it’s made in a Hollywood basement.”
Check out the article online at the Ottawa Citizen or read the print version.