The IBM research lab in Haifa has trained its newest autonomous AI model, Project Debater, to debate complex human issues in front of a live audience.
In a podcast interview published by Scientific American on Wednesday, Project Debater's principal investigator Noam Slonim explained how the AI machine gathers its information for the debate.
The system can analyze upwards of 400 million newspaper articles in just the time it takes to drink a cup of coffee! From that database of information, Project Debater fishes out texts that discuss its desired topic, are "argumentized in nature" and support its side of the debate.
Programmed into the system is also a "collection of more principled arguments that try to capture the commonalities between the many different debates that humans are having," he said. "It looks for the most relevant principle arguments."
An example of such would be the "black market argument," which claims that banning an item could lead to its illegal production. Project Debater could apply this argument to many debates regarding the outlawing of various goods.
Project Debater first showed off its capabilities in 2019, when it went up against award-winning debater Harish Natarajan. At the San Francisco showcase, Project Debater welcomed its opponent in a female voice, "Greetings Harish. I heard you hold the world record in debate competition wins against humans, but I suspect you've never debated a machine. Welcome to the future."
While Project Debater fell short of a win in the event, many were impressed by its human-like communication abilities and its ability to better educate the audience than its human counterpart. In a debate led by Slonim and IBM researcher Ranit Aharonov on the topic of telemedicine, the machine even managed to sway nine people to its side, gaining the majority of the audience.
DARIO GIL, IBM's director of research, informed the audience that the project is "not about winning or losing, but really about the ability to create AI that can master the infinitely complex and rich world of human language."
Dating back to ancient Greece, language rhetoric and debate have been a celebrated and perfected art. Aristotle deemed the art of persuasion to rest upon the three fundamental pillars of Ethos, Pathos and Logos – appealing to authority, emotion and logic.
Ethos refers to one's creditability to speak on the topic. Pathos is the emotional strings a debater pulls. Logos is the factual evidence upon which the argument is based.
Project Debater demonstrated its superiority in researching the facts and the IBM team has been working tirelessly to improve its touch of human emotion. During the San Francisco debate, the machine successfully displayed all three modes of persuasion.
Logos - On the topic of preschool subsidization, Project Debater brought an extensive amount of data and research in favor of the notion: "Statistical summary of studies from 1960 and 2013 by the National Institute for Early Education Research found that high-quality preschools can create long term academic and social benefit for individuals and society..."
Pathos - The system personally expressed, "While I cannot experience poverty directly and have no complaints concerning my own standards of living, I still have the following to share: Regarding poverty, research clearly shows that a good preschool can help kids overcome the disadvantages often associated with poverty."
Ethos - To bolster its credibility, Project Debater began its debate by establishing its knowledge and understanding of the topic at hand and the related lexicon. The system included itself within a population whose moral imperative should be helping better society through supporting preschools."I will argue that we should subsidize preschools," it said. "We are going to talk about financial issues... we accept that the question of subsidies goes beyond money and touches on social, political and moral issues. When we subsidize preschools and the like, we are making good use of government money, because they carry benefits for society as a whole. it is our duty to support them."
Gil excitedly announced at the debate that such progress in the machine's human-like relatability "can really tell us about human thought and expression – and it's this world that is most interesting to us at IBM research. We believe there is great potential in having artificial intelligence that can understand us. The more transparent and explainable we can make AI, the more we can trust it – and the more we can rely on it to help us make better decisions."