In the debate over the virtues of artificial intelligence, get ready for the machines to start chiming in.
On Monday, IBM’s new artificial intelligence system engaged in the first live public debate between a machine and human debaters. It was two Israeli student debating champions and it demonstrated a computer’s ability to deliver persuasive arguments on complex topics in real time.
The system, named Project Debater, was unveiled at IBM’s Watson West site in San Francisco. The arguments were about whether governments should subsidize space exploration and whether the use of telemedicine should be increased.
Preparation for Project Debater began six years ago at IBM’s Haifa research lab, but Noam Slonim, an IBM researcher who created Project Debater, said the AI system only got good enough to debate people two years ago.
During the first debate, the computer was pitted against 2016 Israeli national debate champion Noa Ovadia. Both sides delivered a four-minute opening statement, a four-minute rebuttal, and a two-minute summary in their arguments for each topic.
Project Debater delivered its opening argument in support of government-subsidized space exploration fetching evidence from its massive database, comprising mainly news articles and journals. After listening to Ovadia’s opposing argument, the computer delivered a rebuttal speech. After closing summaries from each side, results from a snap poll showed that the majority of audience members thought Project Debater enriched their knowledge more than its human opponent had.
The second debate, between the computer and Israeli debate expert Dan Zafir, was about whether we should increase the use of telemedicine.
IBM said that Project Debater had not been trained on the topics and the debates were unscripted, except for the jokes which were added during the debates and greetings delivered at its start.
So who won the debates? The crowd judged that the AI system outperformed the human debaters in its ability to use a wide range of information in its arguments, but that the people had the edge when it came to delivery.
Project Debater is the latest AI innovation from IBM, whose Deep Blue system took on chess world champion Garry Kasparov in 1997 and whose Watson beat human champions on Jeopardy in 2011. But debating is a different form of competition, because language and persuasiveness matter.
Project Debater’s ability to absorb massive and diverse sets of information and perspectives to help people build persuasive arguments and make well-informed decisions is new territory for AI, IBM’s director of research, Arvind Krishna, said.