IBM is working on an ambitious, albeit very exciting platform called ‘Project Debater – Speech by Crowd’ for the grand purpose of arguing with people.
Defined as an experimental “cloud-based AI platform for crowdsourcing decision support”, it works by collecting arguments in favor and against a particular topic and using them to build up a debate speech. IBM has stated that it is still in a work in progress. However, their goal is clear: to design an artificially intelligent system that can engage a decent debater on any topic without bias or opinions.
It is important to note that ‘Speech by Crowd’ is a subset of this ultimate project to create a debater fit to argue logically against humans. Instead of absorbing information from vast collections of literature, it builds a debate by crowdsourcing human opinions.
This was demonstrated at CES 2019 where arguments for and against the topic “We should adopt vegetarianism” were taken from the people present over there as well online form around the world. These arguments were subsequently used by the system to create two opposing speeches, which you can experience over here.
In an attempt to show that the AI system hasn’t yet mastered our language, IBM decided to leave the errors in the speeches intact. Therefore any typos that you see in the text have been committed by the system itself.
One of the topics for which arguments were crowdsourced (Image Source: The Next Web)
There is clearly an exciting plethora of applications for such an ingenious system, from coaching a school debate team to helping a lawyer build her case. Dr. Talia Gershon at IBM Research exulted about the performance of the system in an interview: “I was astonished by the persuasiveness of a speech constructed entirely from snippets of ideas collected from people all over the world, and based on discussions at the booth I know I was not the only one.”
All that is left to see now is whether Project Debater’s superior pattern-recognition and analytical skills will be able to completely outwit the best among us.
