Machine Learning. The recipe for more sameness?

Kris Hoet
2 min readOct 24, 2017

I never was a big fan of the much lauded ‘Next Rembrandt’. It was first presented to me in the 2016 Cannes Lions Cyber Jury and apart from more general thoughts around the fact that it misses a link with ING etc, my main problem with it was this: it was an average piece of work, literally. The agency used machine learning and analysed famous Rembrandt paintings to create a new one — it basically took all the averages and maybe that that the measure for the new piece of work. And I thought that was just uninspiring. Incredible achievement from the agency that made it happen and eventually probably successful for ING even but why would the world need an average painting of one of the greatest artists that ever lived? ‘Dreams of Dali’ was the exact opposite of that and therefore way more inspirational. Anyway.

Adobe Scribbler

It makes me wonder if this is where creativity is going. And some of the latest Adobe tools reminded me of that. Because looking at the new tools Adobe just announced that’s not such a crazy idea. Adobe’s Scribbler tool will colorise any of your black & white drawings. How will it do that? By applying it’s knowledge of coloured drawings and making so that yours looks the same. Watson made a movie trailer for a horror movie, based on the analysis of a series of previously made horror movie trailers. And made something that looked as similar as possible to the average of that. When was the last time a creative wanted his/her work to look as much as possible like the average of everything that had ever been done before?

I believe there’s a treasure trove of creative inspiration to be found in behavioural data. And that machine learning can help us guide through all of that at scale. But right now it looks like we’re using that power to create more sameness, rather than something exponentially different which is what it should be.

My 2 cents.