by Mark Waghorn, South West News Service
HAIFA, Israel — A supercomputer has held its own in live debates with humans, openings the door to artificial intelligence taking decision making roles. The remarkable feat will raise concerns about the rise of machines – that can think.
Named “Project Debater,” the ambitious device can discuss about 100 topics – ranging from subsidizing preschool to space exploration and telemedicine. It scans through an archive of 400 million newspaper articles and Wikipedia pages to form opening statements and counter-arguments.
Judged blind by a virtual audience who were given transcripts of the exchanges, it scored highly – against champion debaters.
“The demonstration suggests AI may have the ability to participate in complex human activities,” Dr. Noam Slonim, of the IBM Research Lab in Israel, told South West News Service. “AI makes it possible to produce machines that can perform human tasks. The autonomous system can debate with humans in a meaningful way.”
For some, though, it will conjure darker images: the concept of humans being trapped or enslaved by their own technology. Hal, the murderous AI in Stanley Kubrick’s 1968 philosophical masterpiece “2001: A Space Odyssey” remains the definitive metaphor. But the researchers say there is huge potential in AI that can understand humans.
“We want to show how AI can help make more informed decisions,” said Dr. Slonim. “We all have our own personal biases, based on what we know and our
experience and this can sometimes impact the decisions we make for better or worse. Whether you are a CEO or the mayor of a city having an AI system that can point out your bias with facts and evidence would probably help. You may come to the same conclusion, but at least you are more informed.
“Imagine the Mayor of London considering more green space in the city. One approach could be to ask everyone to vote,” he continued. “But perhaps a more informed approach would be to collect the arguments pro and con from all citizens who might be impacted by the decision. With Project Debater technology we can do that and then summarize numerous arguments into their most prominent key-points – along with their prevalence – for the mayor to consider.”
Ironically the large, rectangular machine is a dead ringer for the mysterious black monolith from the film, apart from a blue animated “mouth.” Speaking in a confident female voice, it opens a discussion with a four-minute speech about a subject from its repertoire – to which a human responds. It then reacts by delivering a second speech of the same length, and the opponent replies with their own four-minute rebuttal. The contest concludes with both participants giving a two-minute closing statement.
“We defined a debate format which is a simplified version of the parliamentary debate style commonly used in academic competitive debates,” said Dr. Slonim. “Once the resolution – called the ‘debate motion’ – is announced, each side has 15 minutes of preparation time.”
In a series of outings, Project Debater took on three of the world’s best debaters. It was unaware of the topics beforehand. Afterwards, the human audiences decided it gave “a decent performance” against Israeli champions Noa Ovadia, Dan Zafrir and Harish Natarajan.
The machine was found to be better in terms of the amount of information it conveyed. In the debate about telemedicine, it was voted to be more persuasive – in changing opinion – than its human opponent Zafrir.
“The challenge seems to reside outside the AI comfort zone, in a territory where humans still prevail, and where many questions are yet to be answered,” said Dr. Slonim.
In Kubrik’s seminal movie, made more than half a century ago, HAL goes mad, declaring: “I’m sorry, Dave. I’m afraid I can’t do that.” With those nine words, the Jupiter-bound Discovery One’s computer intoned a mantra for the digital age.
But Dr. Slonim said many AI algorithms are already involved in decision processes – such as recommending a book or movie online.
He added: “Importantly, this process should be fully transparent, enabling anyone with interest to have a clear look under the covers of the algorithms being used. Correspondingly, all of the AI that IBM designs is build around an ethics policy.”
The U.S. computer giant has spent almost a decade developing Project Debater’s capabilities. They include data-driven speech writing and delivery, listening comprehension for identifying key claims, and formulating principled arguments.
Professor Chris Reed, of the Centre for Argument Technology at the University of Dundee, believes it could help police the internet, including fake news. “Its successes offer a tantalizing glimpse of how an AI system could work with the web of arguments humans interpret with such apparent ease,” he said. “Given the wildfires of fake news, the polarization of public opinion and the ubiquity of lazy reasoning, that ease belies an urgent need for humans to be supported in creating, processing, navigating and sharing complex arguments – support that AI might be able to supply.”
IBM spokesman Chris Sciacca added: “In general, machines that can think are only in Hollywood and science fiction.”