Don’t Worry, IBM’s AI Debater Can’t Beat Humans … Yet

IBM’s AI — the black robot on the left — tried, and failed, to win a debate against humans. Photo: IBM

On Tuesday, IBM’s Project Debater, an artificial intelligence designed to debate humans, attempted to out-argue human debaters in front of a crowd of journalists. Neither the human debater nor Project Debater was made aware of the debate subject beforehand. Debate topics included the necessity of income tax, the safety of self-driving cars, and whether or not antibiotics should be used in food production. The journalists in the room judged the debate using two categories: “delivering the argument” and “knowledge enrichment.” Congrats to mankind on its victory. This time, at least.

While Project Debater failed to win against the human, it did stick up for humans on the issue of mass surveillance. (Even though mass-surveillance technologies like predictive policing deploy AI technology to surveil people). “In the discussion today, I support the people’s right to privacy,” Project Debater said, its voice a strange mix between Alexa and HAL from 2001: A Space Odyssey. “No matter who they are, people should be protected. Mass surveillance is significantly damaging for free expression and the free flow of information around the world.”

The almost-human-size hunk of black metal is designed to “construct well-structured speech on a given topic, deliver it with clarity and purpose and rebuts its opponent,” according to IBM. Eventually, the company wants its AI to help humans by offering “evidence based arguments and limiting the influence of emotion, bias, or ambiguity.” For example, Project Debater could offer pro and con arguments on public policies. (IBM didn’t say anything on how Project Debater might decide what is a “pro” and what is a “con.” Sounds good in theory, though.)

This is not the first attempt by IBM to pit man against machine. In 2011, IBM’s Watson shocked the world by defeating the world’s best trivia stars in a game of Jeopardy. (Both Watson and the humans correctly answered the final question — “Who is Bram Stoker?” — but by that point the computer had already outearned his red-blooded competition.) And 14 years before that, IBM’s Deep Blue defeated world chess champion Garry Kasparov.

While Project Debater may not have won like Watson or Deep Blue, its creators are still optimistic. “It’s unprecedented,” Noam Slonim, Project Debater’s principle researcher, told tech site Calcalist. “There’s no system in the world that can do what’s been shown here today.” At least when the singularity arrives, our robot overlords, if Project Debater’s performance this week is indicative of anything, might protect us against mass surveillance. Hard emphasis on might.

Don’t Worry, IBM’s AI Debater Can’t Beat Humans … Yet