At IBM’s Watson West site in San Francisco, Project Debater, “the first AI system that can debate humans on complex topics”, recently debated 2016 Israeli national debate champion, Noa Ovadia. Each debater presented a four-minute opening statement, a four-minute rebuttal, and a two-minute summary arguing for or against the topic, “We should subsidize space exploration.”
Project Debater was started at IBM’s Haifa, Israel research lab six years ago. The ability of Project Debater to actively engage in a discussion with humans, occurred only two years ago. Three AI capabilities were used in developing Project Debater:
“First, data-driven speech writing and delivery; second, listening comprehension that can identify key claims hidden within long continuous spoken language; and third, modelling human dilemmas in a unique knowledge graph to enable principled arguments (read more about the technical details in the 30+ published papers with access to the training data sets here).”
Project Debater started off the debate with an opening argument that space exploration helps advance scientific discoveries and improves young people’s chances to dream big when considering a big universe- despite the fact that Project Debater never trained on the suggested controversial topics. Project Debater is somewhat of a AI pioneer in the area of free-form debate: “AI can meaningfully engage in full debates on complex topics.”
Noa Ovadia’s eloquent arguments included statements concluding that scientific research here on Earth was more important and vital than investing in space exploration activities. Government subsidies therefore would be more instrumental for this purpose.
Project Debater convincingly produced “a rebuttal speech, countering with the view that potential technological and economic benefits from space exploration outweigh other government spending.”
A second debate involved Project Debater and skilled Israeli debate expert, Dan Zafrir emphasized the theme “we should increase the use of telemedicine.” Both debates demonstrate, for the most part, a previously unachieved aspect of AI, mastering language, an obvious challenge for humans and yet another subject of debate.
