IBM’s Project Debater – an Artificial Intelligence (AI) powered system designed specifically for debating – faced off against Harish Natarajan, a grand finalist in 2016’s World Debating Championships in a live debate organized by non-profit Intelligence Squared US.
Natarajan defeated IBM’s Project Debater, but the debate demonstrated the incredibly complex thinking and arguments that the AI is beginning to be able to make.
The AI system, which has been in development since 2012, was designed specifically with debating in mind, to generate coherent and convincing speeches on its own, while also taking opposing arguments and creating rebuttals.
What makes the Project Debater’s feat so incredible is that it does all of this without being connected to the internet. Project Debater uses newspaper and magazine articles from its own database, and the nuances of the opponent’s arguments, to create its arguments and rebuttals.
This debate again Natarajan followed traditional debate format – each side gave a 4 minute opening speech, then each had 4 minutes for a rebuttal, followed by a 2 minute closing argument by each side. The topic of this debate? Whether or not preschools should be subsidized.
Of the Project Debater, Natarajan had this to say: “it was interesting that it could contextualize information and pull details from research” and that combining these skills with humans could be incredibly powerful.
Though Project Debater ultimately lost, it managed to convince 17% of the live audience to support its side, a very big step forward in language and speech AI.