Today’s artificial intelligence (AI) assistants can tell us what a thing is, like the weather, the square root of 2, and whether a tomato is a fruit or a vegetable. But what happens when you want to ask more complex questions, like is it okay to shoot one chicken if it saves another chicken’s life?
That’s what IBM Research’s Project Debater is exploring (not the chicken — the more complex questions). Project Debater is an AI that has been in the works since 2012, and it — you guessed it — debates. In fact, IBM Research has held the first live, public debate between an AI and humans in San Francisco.
The rules of the debate were simple. There was no awareness of the debate topic ahead of time and there were no canned responses. Each side had to give a four-minute introduction, a four-minute rebuttal to each other’s arguments, and a two-minute closing statement.
According to Fortune and The Verge, Project Debater demonstrated uncanny skill going up against the two champion debaters. The AI was able to analyze volumes of data and create relevant answers. It cited sources, argued against the humans’ points, and preempted possible counter- arguments. It also, apparently, made jokes!
Despites its skill, the publications also reported that the AI sometimes rambled on, with some answers read verbatim from articles. Project Debater also needed a few minutes to analyze the humans’ speeches before it could respond.
But the fact that an AI could conversationally respond to human arguments (relatively) on the fly is a remarkable achievement. And it’s not for the sake of creating something to argue with.
An AI that can sift through volumes of data and engage in discussion could help people make better, evidence- based decisions. Instead of simply telling you what’s what, they can engage you in discussion to present new perspectives. Like whether or not to off that chicken.
PICTURES IBM RESEARCH