Applications and virality: a practical example

If by now you still haven’t used the facial recognition technology of How-Old.net to estimate your age and gender, you must be one of a very few people: the page, launched on April 29, has gone viral, particularly over the last week or so, hitting peaks of more than seven hundred simultaneous users.

And don’t worry: despite initial fears, the company has confirmed that the page doesn’t store or share images or personal information of any kind. The application is essentially the result of an experiment to show how two developers could create something in less than a day using Azure. It’s not that surprising from a technological perspective: similar applications have been around for at least three years, or more even, and it’s not that accurate (it took four years off me, despite my gray hair :-) but it offers us the chance to get a different look at ourselves:

• The app’s creators, Corom Thompson and Santosh Balasubramanian, both information management and machine learning engineers, have shared their experiences on their blog. They were expecting around 50 users: at the time of writing their blog they had already clocked up more than 35,000.
• Eason Wang, Bing’s Senior Program Manager, has published a fascinating article in Quora and Forbes about the application about how the app works by using Bing’s image and voice recognition interpretation technology used in its search engine, along with an API that developers can use. It also mentions another article about how this technology is progressing.
• He’s also published a piece in Medium about the ten reasons that, in his opinion, the app went viral: easy to use, it appeals to our narcissism, it can be surprising and fun, images recall memories and can be shared, it has a good hashtag and web address, it’s simple and does just one thing, there are no language barriers, and was spread through a group of enthusiastic Microsoft employees. I would add what I always tell my students when we’re analyzing these kinds of things: virality still doesn’t actually exist or can be defined as such, and is instead something that simply happens under certain circumstances, but understanding these circumstances is something that definitely helps when it comes to searching it out.
• And finally, Andrej Karpathy, a PhD student in machine learning at Standor, writes another fascinating entry in his blog on this same topic explaining just how far we still are from computerized vision applications becoming genuinely intelligent.

In short, a great opportunity to understand just what lies behind the development of these technologies, and to familiarize ourselves with the analysis of how things go viral. If you noticed the meme in passing but didn’t bother to stop and look closely at it, then now’s your chance to broaden your horizons by reading these articles.

(En español, aquí)