Debate bot talks a good game but human still has the edge

Remigio Civitarese
Febbraio 14, 2019

But a project from IBM shows that computers are getting quite good at it, too.

An IBM-made machine created to win debates has lost out to a human champion debater from London, in a heated competition of man versus artificial intelligence. To generate its arguments and rebuttals, Project Debater uses newspaper and magazine articles from its own database, and also takes in the nuances of the human opponent's arguments. Another interesting element of this project is that it can not copy arguments from sites like Wikipedia. The debate was moderated by John Donvan, four-time Emmy award victor and host of Intelligence Squared US debate series.

The topic of debate was about whether the USA government should subsidize preschools, with Project Debater arguing in favor and Natarajan arguing against. Asked to argue for the proposition, Project Debater used compelling arguments such as how subsidies can help fight poverty, with information backed by various sources. The debate followed the traditional format: four minutes of opening speech for each, four minutes for rebuttal and two minutes of closing speech.

Project Debater took off in a female voice and underscored the need for subsidising preschools by quoting data and studies and also pointing out how it could reduce the crime rate.

"There will still be individuals who will be priced out because of the realities of the market", he said.

If that sounds unfair, consider that the human it debated "has learned all his life", said Aya Soffer, vice president of AI technology for IBM, in an interview before the event. Natarajan addressed specific parts of Project Debater's arguments and rebuffed them - such as saying it's unrealistic to expect a government has an unrestricted budget to put toward helpful programs. Project Debater placed its best points during the conclusion, which at times seemed out of order, but convincing as well. It then tries to capture "the gist" of the opponent's arguments before plumbing the depths of its memory for counterpoints. Amazed by Project Debater's capability to pull details from research and contextualise information, Natarajan said "combining its skills with those of a human can be incredibly powerful".

The victory was awarded based on the percentage of the audience that each side managed to convert: 79% of the audience was initially in favor of subsidizing preschool, while only 13% were against. By the end of the event, 62 per cent of the audience agreed and 30 per cent disagreed. Audience members declared their position before the debate started, and the victor would be the debater that shifted the most people over to their side.

Project Debater shows how AI systems have become increasingly flexible in recent years.

Altre relazioniGrafFiotech

Discuti questo articolo

Segui i nostri GIORNALE