In Humans vs AI we take a look at a particular skill and ask "Who is better at this, humans or artificial intelligence?"
From chess to skiing, we will look at the greatest human achievements in the field and then discuss the best AI competitor.
Each week we'll look at a different topic, and you can catch the latest one in our weekly newsletter, so sign up here!
In what has been billed as landmark moment for AI, IBM this week publicly demoed its Project Debater, an AI that it claims can successfully argue with a human. Do the arguments stack up against its degree educated human opponents or is it all a bit Elementary School my dear Watson? Ai Works takes a look
The stated aim of Project Debater is to
IBM has a previous when it comes to pitting humans against it's AI for
" build a system that helps people make evidence-based decisions when the answers aren’t black-and-white."
It was the subject of much discussion and even a documentary (that you should totally give a watch. It's only 17 minutes. )
This was followed up in 2011 when Watson took on and beat two of Jeopardy's greatest champions.
Fresh from that success, in 2012 the team at IBM set about a far loftier goal. Rather than chess or trivia, could an AI beat a human in a nuanced argument? The result was Project Debater, which was put through its paces against two human opponents.
In San Francisco on Tuesday 19th June, Project Debater faced off against pro debaters Noa Ovadia and Dan Zafrir on two topics “we should subsidize space exploration”, followed by “we should increase the use of telemedicine”.
Project debater (left), Noa Ovadia and Dan Zafrir
Both sides got to make opening statements, before having 4 minutes to rebut the arguments of the other
IBM's device was physically represented by a Kubrick-esque black monolith and spoke with a Siri style automated voice.
The data it had access to was a curated source of articles covering around 100 topics, and it had been trained in part by Ovadia herself, who has been competing with the machine over the last couple of months. In much the same way as Google used real life Go champions to train Alpha Go, IBM has been using real-world debaters to hone their AI's rebuttal skills
It was a draw! The audience voted, and while the humans (unsurprisingly) came out on top in terms of delivery, Project Debater was adjudged to have offered more and better information. Given that it was plugged into a database, perhaps that shouldn't surprise us.
What may come as a bit of a shock is that for the 2nd debate, the audience voted in favor of the AI in terms of persuasiveness.
The potential application of an AI that can convince humans using actual data is remarkable: but there was another noteworthy instance during the debate.
The AI cracked a joke
While discussing how humans might be wary of telemedicine because they are technophobes, it quipped
”I am a true believer in the power of technology, as I should be.”
Hats off to IBM, Not only have they created a machine that can convinclingly argue with a human, it can make lame self referential jokes
This writing may soon be out of a job dear reader!
Who do you think won? Have your say on our twitter poll here.