I am sending pouting selfies to a robot. Its AI is well buff
Unfortunately, it can count my laughter lines
Something for the Weekend, Sir? I like to pick roses on a summer’s day and meeting friends. I dearly wish for world peace. I hope to work with children, just as soon as I have completed my doctorate in astrophysics.
Not really, but I am in training. I am about to enter a beauty contest.
At the risk of slipping back into my default double-entendre mode, it’s a lot harder than you might expect. I am having to take more care over my personal grooming, for example, which in my case means filling three bin bags of body hair shavings every week.
Other things are easy, such as walking up and down a stage with one hand on hip, like a demented John Inman. I don’t want that tiara falling off as I descend the steps from my winner’s throne, do I?
The swimsuit chafes like buggery, though.
Thankfully, this particular beauty contest will involve none of the indignities of “being nice” to the judges or, much worse, being asked my favourite colour by Michael Aspel. This is because the contest will be conducted over the internet and judged by robots.
In case you missed it, the first beauty.ai online competition concluded earlier this year, but v2.0 is soon to be unveiled and I want to be sure I’m in with a chance.
Of all places, I learnt about the next iteration of beauty.ai while attending the Re•Work Digital Learning in Healthcare summit last week in London. One of the recurrent themes of the two days was that the proliferation of smartphones and m-health wearables is enhancing the quality of accurate dynamic health information gathering, which gives the volume and density of data that deep learning systems thrive upon.
Amid the keynotes and presentations about using artificial intelligence to identify Bayesian probabilities in big data, naturally my attention was drawn to what is clearly the technological breakthrough of the decade: a smartphone selfie app that tells you how wrinkled your face is.
With crushing predictability, the app is called Rynkl. Let me not dwell on the modern era’s obsession with misspelt branding as I have written more than enough on this topic in the past. Suffice to say that I continue to have difficulty understanding why every new app is obliged to be named like a 1970s Slade album.
Anyway, the Rynkl team is one of those behind the AI tech that underpins the beauty.ai contest. By taking a selfie and entering whatever age and gender you want people to believe you are, you help train the AI to assimilate common features and recognise the variances between all the faces in your category.
The AI then eliminates everyone who fails to live up to its cyber-chromatic perception of human perfection. The rest of us, I suppose, are headed for the camps. Still, we all have lovely bottoms, eh Ted?
The temptation to orchestrate crowd-sourced sabotage is almost overwhelming.
We could upload head-shots of random wildlife from the savanna, for example, or encourage people to take selfies while gurning like Quasimodo. I am quite certain the AI will spot what’s going on pretty quickly and recategorise us as non-combatants alongside all the ugly Facebook duck-faced pouters, reality TV cocksuckers and Donald Trump lookalikes into a trash folder labelled "Les Dawson impersonators".
So what are the robot judges actually looking for? There are the usual things: symmetry of the face, the relative size and regularity of features, the consistency of skin tone and, of course, that all-essential wrinkle detector.
This somewhat suggests that the AI has been programmed by forcing it to read back-issues of Cosmopolitan in dentists’ waiting rooms.
In the future, when robots will be judging full-body beauty contests, we could program them by feeding them with Daily Mail Online’s sidebar of shame. What’s a “well-toned midriff”? An “amazing post-pregnancy bikini body”? Here you go!
The ideal AI is one that becomes self-learning fairly quickly but at some point early on, someone will have told it what a pretty face is. That someone will have used his or her (I strongly suspect it’s almost certainly a "his") own idea of beauty, which for most people is usually influenced by what they see in the mirror every day.
I don’t know about you but when I look in the mirror, I see a Greek god. Mind you, that’s just my tacky bathroom furnishings. Thank you, thank you, I’m here all week.
No, seriously, if I was to hold up a phone to take a selfie, my face looks pretty good on the screen. I’m a handsome guy, I say to myself. Obviously I say this to myself – I wouldn’t go around saying it to other people or, I dunno, doing anything so rash as writing it down and publishing it in a weekly IT column on the internet.
But that’s not reality, is it? The moment I touch the shutter button, the imaging software’s Reality Mode kicks in. In the photo, I have inexplicably acquired four chins. One of my ears has been raised an inch higher than the other. My nose is spotty. My thick forest of hair has been depleted to a clutch of desperate strands tufting up from my shiny pate like I’d just been given a haircut by Ken Dodd.
I’m fairly sure the AI judges work in Reality Mode. Unfortunately, there’s no way of telling since, even though the initial beauty.ai contest has ended and winners got their prizes – AIs being unaware of the concept of irony, winners were handed gift sets of make-up – their photos have not been made public.
Despite all this, what makes the contest so appealing that one of the key ground rules is NO BEARDS. It’s as if the cutting edge of facial recognition technology has determined that hipsters are officially ugly bastards. And there’s no arguing with that.
Alistair Dabbs is a freelance technology tart, juggling IT journalism, editorial training and digital publishing. He apologises for the relentless references to British TV personalities of the 1970s. If you thought media faces were bad now, he can assure you that it was no better back then… although Rod Hull was still worth 20 Will.I.Ams.