The next great leap in the world of AI has come. It is now possible for a human to debate an AI (Artificial Intelligence). We’ve all been in those embarrassing situations when we get into an argument or a debate with someone who really knows the topic and just makes you feel inadequate. That person may have now met their match and it is quite possible that the AI will wipe that smug smile right off of their face. Of course we have all yelled at our computers when something goes wrong but that is not quite a debate and now your computer could eventually respond back to you. Wouldn’t that be something?
Project Debater
In June IBM unveiled Project Debater in San Francisco. It was a low-key affair and the AI took part in club style debate with no awareness of the topic ahead of time for either side. From what was reported the AI held its own. This was a far cry from IBM’s other famous AI initiatives, Watson, which well made the contestants on Jeopardy (including Jeopardy superstar Ken Jennings) probably feel dumb as it annihilated its competition and Deep Blue which defeated chess grandmaster Garry Kasparov in 1997 (after Kasparov defeated it in 1996).
During the debate Project Debater was able to call on data (it has about several hundred million articles and academic journals), analyze it and create relevant points all within the allotted four minutes. It was able to cite sources, pandered to the audience and also even cracked a joke. While the ability for an AI to make a cogent argument is impressive what was the most impressive was its ability to analyze its human opponent’s argument in near-real time and respond. Debater was not connected to the Internet during the debate.
Project Debater’s performance
Now Debater’s performance was certainly no Frank the Tank from Old School (ironically fielding a similar topic to Project Debater!) and it still has a long way to go before it is capable of entering politics and becoming the first AI US President. In a way it would be refreshing. This AI does not play games but instead is analytical. It assigns a confidence value to every piece of data that it analyzes so it uses only information that is known to be correct or has a high probability of being factual. Imagine that, bringing facts to debate!
Now to be fair there were some things that the AI struggled with when it came to debate. It did not address it’s human opponent’s argument directly, choosing instead more broader arguments, but then it’s not like humans don’t do that either. It also took several minutes to digest both the topic and its human opponent’s argument, something that would not be permissible in a real debate. But all in all considering that this is still a research project Debater’s performance was quite impressive for those involved. The humans were credited with having a better delivery but Debater was credited with a more substantial argument, which should not be a shock at all.
Could AI enter politics??
Of course IBM is not intending for one of their AIs to become a world ruler. The project is intended to help AI understand language in the hope that one day it could be used to help a lawyer create an argument or even to remove the scourge of the Internet today: fake news. With the ability to crack a joke and to pander to its audience though one does have to wonder if politics could be in Debater’s future?
Given the dysfunction in our state and national governments one can see the appeal of having AI involved in government in some way. A whole legislative body of AIs could do what it takes our current legislative bodies years to do in just a matter of minutes. It does almost sound strangely appealing but for now we will have to be content with just debating it.