Breaking ground, IBM Haifa team holds live robot debate fed by crowd arguments

The tech, when commercialized, could help companies and governments collect opinions, make more informed decisions

Shoshanna Solomon was The Times of Israel's Startups and Business reporter

IBM's Project Debater discussing pros and cons of legalizing cannabis based on arguments collected from the crowd, June 13, 2019 (Or Kaplan)
IBM's Project Debater discussing pros and cons of legalizing cannabis based on arguments collected from the crowd, June 13, 2019 (Or Kaplan)

In a global first, IBM’s team in Israel held a live event in which a robot called Project Debater was able to collect arguments generated by a crowd, analyze them, create pro and con narratives and deliver them as speeches, all within an hour.

Audience members in a jam-packed hall in the Expo Tel Aviv exhibition grounds were given a link where they were asked to feed arguments in favor of or against legalizing cannabis. The arguments needed to be between eight and 36 words, in English.

Then the robot set to work, meeting a number of challenging tasks: it scoured all of the 1,037 pro and con arguments that had been uploaded and classified them into support or contest fields. Then it removed redundancy; namely, it recognized that people were making the same point using different words.

Third, it managed to automatically detect the underlying themes of the debate that emerged from all contributed arguments and organize the narratives in a coherent manner. And finally, it managed to identify the key points raised by the crowd, both for and against the premise, and lay them out clearly in two pro and con speeches.

“Legalization of cannabis would create businesses that the government can tax,” the robot, a tall black pillar with three round, blue flashing lights facing the crowd, said in a clear, slightly mechanical woman’s voice, stating one of the pro arguments. Or, “cannabis could become a gateway to drugs,” it said in its speech summarizing the cons.

Seventy-six percent of respondents were in favor of legalizing the weed, whereas 24% were against, the attendees were told.

“We are very happy” with how the Project Debater – Speech by Crowd demonstration went, said Dr. Noam Slonim, principal investigator, Project Debater, IBM Research – Haifa, on the sidelines of the event. They were very concerned before the event, he said, that people wouldn’t contribute arguments or that the system would crash.

“But it went very well,” he said, despite the fact that the robot had encountered glitches. For example, he said, the robot put a pro argument in a con space — the idea that legalization of the drug would kill the black market for it — because it heard the words “kill” and “black market,” both of which sound like bad things.

The robot’s speech also had some unnecessary repetitions, he said, sounding like a parent criticizing a child’s performance.

But on the other hand, Slonim added, the fact that there were some errors and repetitions not only proved that the demonstration was really live, but also made the robot seem “more human.”

Dr. Ranit Aharonov, Manager, Project Debater, IBM Research – Haifa, left, with Noam Slonim, Principal Investigator, Project Debater, IBM Research – Haifa, at the demonstration of the crowd-sourced arguments for Project Debater; Tel Aviv, June 13, 2019 (Or Kaplan)

This technology was conceived at IBM’s Haifa research lab and is intended to enhance humans’ decision-making when faced with complex questions.

Slonim has worked closely with project manager Dr. Ranit Aharonov and IBM’s research team on Project Debater since 2012, when it was selected by IBM to be its next grand global challenge.

The project, which is being hailed as a new era for the interaction between humans and machines, has held several demonstrations around the world showcasing the robot’s ability to take on champion human debaters, including a live debate in San Francisco earlier this year.

But this was the first time that the machine was put through its paces using arguments gathered from the crowd in a full live debate, as opposed to gathering the arguments from a database.

The development of the technology is meant to “enhance human decision making,” by processing information quickly, said IBM’s Aharonov on the sidelines of the event.

IBM Israel CEO and country general manager Daniel Melka at an event in Tel Aviv, June 13, 2019 (Or Kaplan)

There could be many practical uses for this technology, said Daniel Melka, country general manager, IBM, also on the sidelines of the event. It could be used by CEOs to gather their employees’ views to help in their decision making process; by companies to get feedback from customers about their products or services; or by government institutions, to get and analyze views from residents about a variety of subjects, he said.

There is no time frame yet for commercialization of a specific debater product, he said. But IBM will be soon making the product available to interested customers, who will get a link into which they can upload feedback and thus help the IBM researchers refine their work to suit a variety of needs.

“This tech could turn into a product around the decision-making process,” Melka said.

IBM has been active in Israel for the past 70 years, and its Haifa team is the largest lab of IBM Research Division outside of the United States. Founded as a small scientific center in 1972, it grew into a lab that leads the development of innovative technological products and cognitive solutions for the IBM corporation. Its various projects utilize artificial intelligence, cloud data services, blockchain, healthcare informatics, image and video analytics, and wearable solutions.

Most Popular
read more: