Personal decision-making in a world of disinformation could become a whole lot easier, thanks to IBM’s new AI technology.
In front of 700 people in a San Francisco theatre, it put a version of its Watson supercomputer up against one of the world’s best debaters. For 25 minutes, man and machine traded arguments over the motion that we should subsidise preschools.
IBM’s Project Debater appeared on stage in the form of a long black tablet, three balls moving on its screen as it spoke in a slightly robotic female voice.
These were not brief answers regurgitated from Wikipedia, as you get from Google or Amazon’s “smart” speakers. For up to four minutes at a time, Project Debater reasoned and argued in favour of the debate resolution, mining 10 billion sentences of reference material to produce a compelling argument for why subsidising preschools helps kids get the best start in life.
As I sat in the audience, it was quite something to behold. IBM claims the only preprogrammed part of its spiel was the opening greeting to its opponent, 31-year-old Oxford- and Cambridge-educated debater Harish Natarajan.
The complexity involved in programming algorithms that can make sense of a question, structure a coherent answer and assemble the facts to back it up is staggering. The computing power required is dozens of times more than is available in your average laptop.
Even more impressive was that Project Debater listened to Natarajan’s clever rhetoric and came back with reasoned replies. But in the end, Natarajan won, with the audience determining via electronic voting that he had the more compelling argument.
Project Debater had facts aplenty, but Natarajan’s arguments were full of what appeals to humans, what Aristotle called pathos.
IBM has pulled these types of stunts before. It pitted its Deep Blue supercomputer against chess grandmaster Garry Kasparov in 1996-97, and its Watson supercomputer competed against top contestants in the quiz show Jeopardy! in 2011.
But this contest was different. Even the Israeli engineers who built Project Debater didn’t seem particularly disappointed that their creation lost. They know it will be incredibly hard for a supercomputer to better the rhetorical skills of a top human debater.
But Project Debater suggests there’s a more nuanced coexistence ahead for humans and machines, one in which artificial intelligence augments our abilities rather than replaces them. That’s also a more tenable position for the tech companies that are pursuing artificial intelligence, but increasingly facing a backlash as concerns grow that they will be responsible for putting millions of people out of work.
The question now is what IBM can do with the incredible mix of artificial intelligence and information gathering that Project Debater represents. Could it be the answer to the fake news and misinformation that continue to clog social media?
What if IBM’s machine was put to work analysing tweets and Facebook posts in real-time, rating them for factual accuracy against a high-quality database of articles and ranking them in our newsfeeds based on the results? We could have Project Debater listen to Parliament TV and deliver a fact-check of our politicians’ claims.
Given the exponential nature of increasing computing power, it won’t be that long before something similar to Project Debater is available to us, as a true smart speaker in the home.
Then we’ll have the chance to ask the oracle in the corner just about anything and receive a rational, fact-based answer. Imagine basing our decisions in life on solid evidence rather than dud advice from friends, or simply a hunch.
With fickle, impulsive humans, Project Debater has her work cut out.
Peter Griffin travelled to San Francisco as a guest of IBM.
This article was first published in the March 2, 2019 issue of the New Zealand Listener.