Microsoft website goes viral by telling people how old they look

A face-detection website that estimates how old you look has gone viral, providing Microsoft with the chance to illustrate the power of its Azure cloud service during Build 2015.
Written by Jack Schofield, Contributor
Paul Allen looks 33 while Bill Gates looks 23, says how-old.net
The HowOldRobot reckons bearded Microsoft co-founder Paul Allen looks 10 years older than Bill Gates. In fact, the age difference was 1,010 days or 33 months.
So you can have a huge demo day and show tons of innovative technology - cross-platform coding, phones that work as PCs, a ground-breaking augmented reality headset etc - but if you really want to set the internet on fire, just offer to tell people how old they look.

That's what Microsoft's Azure services and apps have been doing today, and it turned out to be a good illustration of how its cloud can track virality in real time and scale up to meet demand. And then, no doubt, scale back down in a few hours when the whole thing blows over.

The feeding frenzy started yesterday afternoon (Wednesday, 29 April), with a post on Microsoft's Machine Learning Blog: Fun with ML, Stream Analytics and PowerBI - Observing Virality in Real Time

A couple of software engineers -- Corom Thompson and Santosh Balasubramanian - had been exploring "Microsoft's newly released Face detection API's through a webpage called http://how-old.net."

They emailed a couple of hundred users to get some beta-test feedback, "optimistically hoping that at least 50 people would give it a shot." In a few hours, they had 35,000, albeit around 29,000 were from Turkey. Today, the conflagration spread through Twitter and, as you know, Twitter drives the tech media industry. The result was a rash of blog posts, including this one.

The website is extremely simple: you choose or upload a photo and the #HowOldRobot tells you how old you look. It's obviously better to get a robot to do this than a human being: robots are not too smart, so you won't be too insulted. And because it's saying how old you look, not how old you are, it's never going to get your age wrong. Of course, it might be wrong about how old you look, but there's bound to be some wriggle room.

Or maybe you just have a beard, which can add quite a bit to your age. As our own Ed Bott found.

If I'd been programming how-old.net, I'd have given it a tendency to underestimate the age of anyone over 30. Not many people are going to complain about being flattered, though I expect the under 21s might be flattered if they're told they look slightly older and therefore more mature.

A sink: humans see faces where programs don't
Meanwhile, more cunning souls were probably uploading photos of dolls (hi, Tom Warren), pets, pop stars, and everything else from dead presidents to pareidolian baked potatoes. Happily, trying to fool AI systems is harmless fun, not a matter of life and death. Yet.

Either way, the blog post provides technical information about how it was done, including some very short bits of code. The programmers conclude: "we hope you have fun with it and are inspired to create your own solutions using Azure services and the APIs available in the ML Gallery".

The site's real message is that, with Azure, a couple of programmers can create something like this in a day. The age-guessing is just window-dressing.

Editorial standards