ARCHIVAL (“2001: A SPACE ODYSSEY,” 1968):
ASTRONAUT: Open the pod bay doors, Hal.
HAL: I’m sorry, Dave, I’m afraid I can’t do that.
NARRATION: If you had to pick our favorite fictional story about robots, the one where they wipe out humans keeps on delivering at the box office.
ARCHIVAL (TRAILER FOR “THE TERMINATOR,” 1984):
VOICEOVER: The Terminator.
ARCHIVAL (“THE MATRIX,” 1999):
AGENT SMITH: Human beings are a disease and we are the cure.
ARCHIVAL (“EX MACHINA,” 2015):
MAN: Ava, I said stop!
NARRATION: And recently, similar fears about artificial intelligence seem to be spilling into the news.
ARCHIVAL (BLOOMBERG, 4-18-18):
REPORTER: The robot apocalypse could be closer than you think.
NARRATION: Futuristic as all this sounds, history has some insights to offer us here. Because, even on the news, we’ve seen a version of this movie before. In the 1990’s, the media was fixated on a real life, high-stakes battle between chess grandmaster Garry Kasparov and IBM supercomputer Deep Blue.
MAURICE ASHLEY (CHESS GRANDMASTER): Everybody billed it like this was the terminator come to potentially take down the humans.
ARCHIVAL (ABC EVENING NEWS, 5-3-97):
REPORTER: IBM’s 3,000-pound supercomputer, which can calculate 200 million chess positions per second.
ARCHIVAL (CNN, EVENING NEWS, 5-9-97):
REPORTER: All the major TV networks have covered it, and it’s been beamed to 20 countries around the world.
MAURICE ASHLEY: I was rooting for Kasparov to kick its ass. There’s no question about it.
NARRATION: While chess grandmaster Maurice Ashley was giving live commentary on the match, Murray Campbell was rooting for Deep Blue. He helped design it.
MURRAY CAMPBELL (SENIOR MANAGER, ARTIFICIAL INTELLIGENCE, IBM): Chess was commonly considered to be a grand challenge for computer science. The earliest computer scientists said, ‘If we can get a computer to play chess we’ve really done something.’
NARRATION: And in the first round Kasparov had the upper hand.
ARCHIVAL (CNN EVENING NEWS, 5-12-97):
GARY TUCHMAN: He won game one versus the Deep Blue supercomputer.
COMMENTATOR: In fantastic style.
NARRATION: But Game Two changed everything. About 35 moves in, Kasparov set a trap. But Deep Blue refused the bait. Instead, the machine made a shrewder choice, that paved the way to a win.
MAURICE ASHLEY: It was stunning to see a computer play like that. When you have a choice between an aggressive, sharp, tactical move that is concrete and specific, versus a subtle positional move, that’s really where the-the grandmaster is shown.
MURRAY CAMPBELL: Those sequence of moves showed Kasparov that Deep Blue was playing at a level beyond what he had imagined it could do.
NARRATION: A shaken Kasparov resigned about 10 moves later. In the rest of the games, Kasparov fought to a grueling series of draws until, in the sixth and final face-off, the exhausted human champ fell apart completely.
ARCHIVAL (ABC EVENING NEWS, 5-11-97):
MAURICE ASHLEY: There was no reason for him to play chess like this. He never plays chess like this.
ARCHIVAL (CNN EVENING NEWS, 5-11-97):
CYNTHIA TORNQUIST: He resigned about an hour and three minutes into the game.
ARCHIVAL (ABC EVENING NEWS, 5-11-97):
GARRY KASPAROV: I have to apologize again. I am ashamed by what I did at the end of this match.
NARRATION: Media pronouncements on the outcome’s gloomy implications were swift.
ARCHIVAL (CBS EVENING NEWS, 5-12-97):
DAN RATHER: We humans are trying to figure out our next move.
ARCHIVAL (CBS EVENING NEWS, 5-11-97):
TROY ROBERTS: Call it a blow against humanity.
ARCHIVAL (ABC Evening News, 5-12-97):
JEFF GREENFIELD: The victory seemed to raise all those old fears of superhuman machines crushing the human spirit.
NARRATION: But computer scientists had a different reaction.
PATRICK HENRY WINSTON (PROFESSOR OF ARTIFICIAL INTELLIGENCE, MASSACHUSETTS INSTITUTE OF TECHNOLOGY): Every time a computer does some narrow thing better than a person there’s a temptation to think that it’s all over for us. But Deep Blue doesn’t play chess the way Kasparov plays chess. Deep Blue processes information much like a bulldozer processes gravel.
GURUDUTH S. BANAVAR (CHIEF TECHNOLOGY OFFICER, VIOME): Every slice of capability that we’ve seen computers become really good at, and even superhuman at, are actually one small sort of small pieces of the breadth of intelligent behaviors that we exhibit.
NARRATION: Guru Banavar helped build the digital descendant of Deep Blue – Watson. It’s a talking, self-teaching system, nimble enough to play Jeopardy. In fact, it became very hard to beat.
ARCHIVAL (ABC, WORLD NEWS WITH DIANE SAWYER, 2-15-11):
WATSON: Who is Michael Phelps?
ALEX TREBEK: Yes. Watson?
WATSON: What is Last Judgement?
NARRATION: So how close are machines coming to outsmarting mankind? The people working to solve some of AI’s toughest problems may be in a unique position to know. For example, before smart machines could run amok, they’ll need to walk. At MIT in 2016, Russ Tedrake led a team of engineers designing software for one of the most advanced humanoid robots ever built.
RUSS TEDRAKE (PROFESSOR OF ELECTRICAL ENGINEERING, MASSACHUSETTS INSTITUTE OF TECHNOLOGY): The level of complexity that we can deal with is absolutely state-of-the-art and beyond.
NARRATION: And if machines are going to walk, they’ll need to recognize what’s in front of them. A few years ago at Stanford, Fei-Fei Li taught computer systems to describe objects they see in pictures for the first time.
ARCHIVAL (FEI-FEI LI TED TALK, MARCH 2015):
COMPUTER COMPUTER VOICE: A man is standing next to an elephant. A large airplane sitting on top of an airport runway.
FEI-FEI LI (PROFESSOR OF COMPUTER SCIENCE, STANFORD UNIVERSITY): We’re really on the quest for building machines and computers to have that kind of visual intelligence that eventually can match to humans. Visual intelligence is about seeing the objects, understanding the scene, reasoning about the visual story.
NARRATION: At MIT, Patrick Henry Winston has been programming systems to carry out the kind of basic reasoning people use to interpret stories.
PATRICK HENRY WINSTON: What is it that makes human intelligence different from the intelligence of something like a chimpanzee, or a Neanderthal? And for me it’s the ability to tell stories.
NARRATION: Each of these scientists’ projects amounts to an engineering moonshot in its own right. Yet each aims to replicate just one facet of the general intelligence humans take for granted. And even as the technology improves, none of these researchers see a finish line in view.
RUSS TEDRAKE: This is absolutely one of those very state-of-the-art-machines. But it is not capable of even some of the things that we’d expect a toddler to be able to do very effectively.
FEI-FEI LI: I’m not trying to say we didn’t work hard, and you know, we have made a lot of progress. But I think it’s important to understand we are closer to a washing machine than the Terminator.
ARCHIVAL (FEI-FEI LI TED TALK, MARCH 2015):
COMPUTER COMPUTER VOICE (DESCRIBING A PHOTO OF A STATUE AND ANOTHER OF A BABY HOLDING A TOOTHBRUSH): A man riding a horse down a street next to a building. A young boy is holding a baseball bat.
PATRICK HENRY WINSTON: The closer you come to doing research in this area the more you realize how difficult everything is. We don’t know when those discoveries will come. But they look like there’s going to be many of them, not just one.
NARRATION: And these scientists say it’s unlikely we’ll see smart machines beget vastly smarter versions of themselves overnight and totally escape human control. That’s because these AI nightmare scenarios fail to grasp a paradox that underlies much of the work in artificial intelligence.
GURU BANAVAR: Things that are easy for humans are hard for computers. And things that are easy for computers are hard for humans. We underestimate all of the things that we do so easily.
NARRATION: In some ways, it comes down to common sense. We see this problem in one of the most visible applications of AI on the street right now. Cars owned by the Google offshoot Waymo are piloting themselves around a suburb of Phoenix, Arizona, as part of an experimental driverless taxi service. It works in part because Waymo cars follow hyper-detailed maps.
ANDREW CHATHAM (LEAD SOFTWARE ENGINEER, MAPPING, WAYMO): Our maps have, down to about 15 centimeters, the location of every curb, traffic light, stop sign, driveway. And so, for a car from us to appear on your block, we need to have built a map of your block.
NARRATION: The question is what happens in more chaotic situations that call for more common sense understanding on the road?
JOHN LEONARD (PROFESSOR OF MECHANICAL ENGINEERING, MASSACHUSETTS INSTITUTE OF TECHNOLOGY): There’s this, sort of, negotiation, which has been called the social ballet of driving. How do you write the computer code that says, always stop at red lights unless there’s a man on the side of the road who’s a police officer and is waving you to go through a red light. That’s a really hard thing to do.
NARRATION: Obeying a traffic cop is just one common sense task humans carry out behind the wheel, and things like this remain hard for machines. And that’s why Waymos aren’t likely to appear soon on your block if the conditions aren’t ideal. AI works best on problems where there’s a structured environment.
While some researchers worry that millions of workers could be displaced by automation, others think our jobs will simply be transformed. And one of the optimists on this issue may surprise you, Gary Kasparov.
ARCHIVAL (TED TALK, 2017):
GARY KASPAROV: Human plus machine isn’t the future, it’s the present. And as someone who fought machines and lost, I am here to tell you this is excellent, excellent news.
NARRATION: As for the question about Hollywood fears…
ARCHIVAL (“AVENGERS: AGE OF ULTRON,” 2015):
ULTRON: I’m glad you asked that, because I wanted to take this time to explain my evil plan.
NARRATION: …plenty of AI researchers say we’re safe from those for now.
RUSS TEDRAKE: I think you can’t watch this robot without thinking, “Wow, they’ve got a long way to go.” We like to joke, “His batteries only last an hour, so, you know, even if he ran amok he wouldn’t get very far.”