NEXT: Will Supercomputers Ever Deliver TED Talks?
If you haven’t seen Ken Jennings’s 2013 TEDxSeattleU talk, you owe it to yourself to watch.
Jennings, you’ll remember, was the undisputed champion of Jeopardy, until he and Brad Rutter were roundly defeated by Watson, the massively parallel processing computer developed by IBM. Even more impressive than the way Deep Blue — another of IBM’s creations — defeated Garry Kasparov, the undisputed champion of chess, Watson dispatched with the two human competitors without so much as breaking a digital sweat.
As he revealed his Final Jeopardy answer — which he got correct, but to no avail as Watson had an insurmountable lead — Jennings also wrote “I, for one, welcome our new computer overlords.” With that one line, Jennings was doing two things. First, he was commenting on the meaning of the event that had just transpired: that computer intelligence is quickly approaching that of human intelligence such that the machines will soon be smarter than we are. At the same time, Jennings was paraphrasing a quote from Kent Brockman, the news anchorman on the Simpsons. You may recall the episode where Homer is launched in space, and in his bumbling fashion, smashes open an ant colony. One of the ants floats in front of a live camera making it appear enormous. Brockman, shaken by what he has seen, pronounces the earth is under attack by giant insects, and is also quick to sycophantically add “I, for one, welcome our new insect overlords.”
With that one line scribbled on the Final Jeopardy screen, Jennings had simultaneously made sense of events and had made a pop culture allusion.
And then he gave a TED talk. In the talk, Jennings describes having “the future come right at him,” as he waxes philosophical on the meaning of Watson’s victory. He also makes a plea for the preservation of human knowledge at a moment when synthetic intelligence seems to be irrevocably destined to triumph. He seeks those who will “lead the future” as we develop smarter and smarter machines capable of performing feats of intelligence much more complex than playing a trivia game.
I think Jennings is wrong in some places, and I would welcome the chance to debate him on some of these points, especially the kinds of human intelligence that will not be replicated or supplanted by synthetic intelligence.
But this is precisely the point: Watson cannot do any of the things that Jennings can do. Watson can win at Jeopardy and is now used to, among other things, aid physicians in making diagnoses. Watson is indeed intelligent, but has yet to comment on the meaning of its victory, and cannot make pop culture or literary allusions. Watson has yet to give a TED talk. Nor could I engage in a debate or conversation with Watson about any of this.
Computers and algorithms acting on big data appear poised to develop a kind of superintelligence, and many thoughtful commentators are imagining a world where synthetic intelligence has made more and more humans redundant. (Such a world could be a reality within the next twenty years.) When machines replaced human labor, economists and technologists said that humans were now free to engage in “smart jobs” or those occupation that require cognitive skill. Now even those cognitive jobs no longer appear safe, as companies like IBM and others develop algorithms that work as well as — and faster and cheaper — than human beings. Paralegals and accountants and travel agents have already felt these effects.
The “humans made redundant by synthetic intelligence” scenario is plausible only if Moore’s Law continues to hold: that computational speed will continue to double every 18-24 months. There are some observers, however, who suggest that at some stage the laws of physics will intervene, and our ability to manufacture such exponentially increasing computing power will reach a limit. That limit may fall well short of the superintelligence some fear. More plausible to me is the “humans will cooperate with very intelligent machines to do things neither could do alone” scenario, where synthetic intelligence can reveal patterns in large data sets, and where their human partners will make sense of and unravel the meaning of those patterns.
We will indeed continue to develop intelligent computers, but I, for one, do not foresee a day when they become our overlords. Only when machines are able to engage in sense-making, to interpret, to make literary allusions, to engage in real conversation will they begin to approach the level of intelligence practiced by humans.
We’ll know that moment has come if and when Watson ever gives a stirring TED talk.
David Staley is president of Columbus Futurists and a professor of history and design at The Ohio State University.
The next Columbus Futurists monthly forum will be Thursday January 21 at 6:30 PM at the Panera Bread community room (875 Bethel Rd.) Our topic for the evening will be “Predictions for 2016”.