NICK EICHER, HOST: Today is Thursday, June 28th. Thank you for turning to WORLD Radio to help start your day.
Good morning. I’m Nick Eicher.
MARY REICHARD, HOST: And I’m Mary Reichard.
Over the years, IBM has made some spectacular strides in advancing the frontiers of artificial intelligence. You might recall that IBM’s Deep Blue system defeated world chess champion Garry Kasparov back in 1997. And who can forget how IBM’s Watson trounced two top Jeopardy! champions in 2011?
EICHER: Following the strict rules of a game like chess or quickly answering trivia questions is one thing. But what about formulating and articulating complex arguments? IBM may have pulled a hat trick with its latest innovation: a computer that can debate.
At an event earlier this month in San Francisco, IBM’s newest AI system, Project Debater, took on two champion human debaters in front of a live audience who came away impressed by its persuasive abilities.
REICHARD: WORLD Radio technology reporter Michael Cochrane is here now to tell us more about robot debaters and what that might mean.
Michael, first of all, describe the debate setting for us: What was the format and what were the topics?
MICHAEL COCHRANE, TECHNOLOGY REPORTER: Sure. The first proposition was the statement, “we should subsidize space exploration.” Project Debater opened with four-minute argument supporting the statement. Then Noa Ovadia, the 2016 Israeli national debate champion, had four minutes to oppose the statement. Project Debater listened to Ovadia’s argument and then gave a 4-minute rebuttal speech. Following 2-minute closing summaries from both sides, a poll of audience members revealed that a majority thought the computer had done a better job of increasing their knowledge of the subject.
Wow, so Project Debater had to essentially engage in unscripted, persuasive reasoning?
COCHRANE: Exactly! That’s what’s so revolutionary about this. Project Debater was trained on debate methods but only found out what the topic was moments before the debate started, unlike its human opponent. It did have access to a database of 300 million curated news articles and scholarly papers, but it still had to figure out what material supported its argument, make a persuasive case, listen to its opponents’ arguments and then rebut them. It even demonstrated what’s called, procatalepsis, which is the ability to anticipate your opponent’s argument and engage it before they do.
That’s pretty amazing! How did IBM physically represent Project Debater during the debate?
COCHRANE: IBM made Project Debater into a tall, black obelisk on the stage—clearly paying homage to science fiction writer Arthur C. Clarke’s “2001: A Space Odyssey.” The AI even spoke with a well-modulated female voice, kind of evoking the sci-fi movie’s famous HAL 9000 computer. Listen to a clip of Project Debater closing out its rebuttal speech and see if you can’t appreciate how this system is able to draw on a broad range of persuasive skills:
HAL: Having a space exploration program is a critical part of being a great power. Investing in space exploration will bring positive returns for the country. If I may put it a bit differently: subsidizing space exploration is like investing in really good tires. It may not be fun to spend the extra money, but ultimately you know both you and everyone else on the road will be better off. For all of these reasons, I think the motion should stand. We should subsidize space exploration. That concludes my speech. Thanks for listening.”
Impressive! And maybe a little scary, too. So, why did IBM invent this type of AI? What was the goal of the research?
COCHRANE: IBM started work on Project Debater in 2011 with the goal of developing AI that learns across different disciplines in order to augment human intelligence. They were particularly concerned with mastering language. Specifically, they gave Project Debater three capabilities: data-driven speech-writing and delivery, secondly, enhanced listening comprehension, and finally, the ability to model human dilemmas to support principled argumentation.
A system like this might be useful for helping people make better decisions, then?
COCHRANE: That’s exactly right. IBM thinks Project Debater or a system like it could become the—quote—“ultimate fact-based sounding board without the same bias that often comes from humans.”
Let’s talk about that bias for a moment. Can you think of an example where an AI like this could have “moved the needle” so to speak on a policy issue—removing the human bias?
COCHRANE: I think it would have been very helpful to have had a tool like this during the debate over healthcare reform, for example. So much of what passes for “debate” these days on hot-button issues is highly politicized and ideological. An AI like this could really help humans evaluate policies and courses of action in a rational, dispassionate way.
With its huge knowledge base and ability to form and defend unbiased arguments, this could help cool down much of the rhetoric surrounding policy decisions, right?
COCHRANE: I agree. And I can even imagine a scenario in which you could, for example, pit one AI debater against another AI… then stand back and see what convincing arguments emerge from the debate. That would be awesome to watch!
So, Michael, if AIs like this become accessible to ordinary people like us, what kind of applications can you foresee?
COCHRANE: I could see widespread use of this kind of tool in school settings – including homeschoolers – with the goal of teaching kids real critical thinking skills and the ability to train their minds for true argumentation and debate; not simply shouting a list of talking points at an opponent.
And maybe replace lawyers someday, who knows? Michael Cochrane is our science and technology reporter. Thanks for this report, Michael.
COCHRANE: You’re welcome, Mary.