People are usually good at arguing, and the ability to comprehend and put forward arguments is one of our greatest capabilities. But what about AI computers that try to emulate humans?
This is precisely what IBM wanted to find out by pitting its AI system dubbed the Project Debater with world-class debater Harish Natarajan. The topic for debate was — “whether or not preschool should be subsidized” with Project Debater arguing in favor and Natarajan arguing against it.
Even though Natarajan won the challenge, but the artificially intelligent computer exhibited highly complex arguments as proof of progress in AI.
Natarajan is the 2016 World Debating Championships Grand Finalist and 2012 European Debate Champion, whereas Project Debater is IBM’s homebaked system designed to come up with logical and coherent speeches of its own.
The Debater is capable of taking in the arguments of a human opponent and creating its own responses as well as closing argument. To come up with arguments and rebuttals, the system utilizes information gathered from newspaper and magazine articles from its own database.
What’s more impressive is that fact that it can adapt to the nuances of the human opponent’s arguments. Moreover, it isn’t connected to the internet and hence cannot pick up arguments or stats from sites like Wikipedia.
The topic of the debate wasn’t revealed to the AI system or Natarajan until 15 minutes before the debate started. Before the debate, 79% of the audience agreed that preschool should be subsidized and only 13% disagreed.
However, at the end of the debate, 62% of the crowd agreed and 30% disagreed. In this style of debate, points are measured by which side gains the most percentage points, hence Natarajan took the win.