IBM’s Watson Predicts Cannes success for MullenLowe London Posted on

Can Watson pick a Cannes Lions winner? IBM’s cognitive system tries its arm at judging awards

Judging awards is a cognitive process, so it should be elementary to dear Watson. That is why we gave IBM’s cognitive system the ultimate creative test.

By ingesting 1000s of previous Cannes winners and losers, Watson has learned to be one of the best informed jurors in the business.

Now, we have asked it to take a look at the shortlisted work in the Outdoor category before the winners are announced today and tell us whether any UK or US agencies stand a chance.

Here are Watson’s favourites…

  • UK Gold: MullenLowe London, Persil – Free the Kids
  • UK Silver : J. Walter Thompson London, Listerine – Where’s your mouth been?
  • UK Bronze: McCann London, Ethos – Time to get away
  • US Gold: TBWA\Chiat\Day New York, Airbnb – Belong Anywhere
  • US Silver: Energy BBDO Chicago, Raid – ‘Death has never been more delicious’
  • US Bronze: Energy BBDO Chicago, Ziploc – ‘Life needs Ziploc’

How we did it…

Oliver Cox, solutions architect at IBM Watson Ecosystem explained: “More than a thousand previous Cannes Lions winners and losers from the outdoor category (an equal number of each) were randomly split these 80/20, with the 80 per cent used to train Watson to create a classification model that could tell the difference between a winning poster campaign and a losing poster campaign, and the remaining 20 per cent used to test the classification model.

“Since it was already known whether these had won or lost, we were able to measure its accuracy and in the last test we did 90 per cent of losers were classified correctly and about 78 per cent of winners, meaning it is pretty accurate.”

The Drum’s magazine editor Thomas O’Neill added: “This is an experiment that could massively disrupt the awards industry. We have the potential here of AI being able to identify an award winning ad from a loser before you’ve even bothered splashing out on the entry fee. We’re looking forward to seeing whether it proves as accurate in reality as it did in training.”

This article was originally published on The Drum