india

Faking it well

A website explores ways to cheat facial recognition algorithms. It's good training for impending dystopia.

Think of it as training, not a game. If the science fiction dystopias that flood OTT platforms — and the ever-growing dangers of social media — have taught us anything, it is that the machines will take over. Already, facial recognition software are deployed on a vast scale — from job interviews to casinos and even to predict crime, a la Minority Report — with algorithms that claim to be able to “read” the emotional state of people. In Lucknow, for example, there were reports of a planned pilot project that will set AI-enabled cameras at “harassment hotspots”. If the expression of a woman changes to one of distress, the nearest police station will be alerted. Now, a team of scientists has set up a website that allows users to explore the kinds of facial recognition software out there and try to beat them.

On emojify.info, people can play a series of games, watch videos and leave their comments. Apart from spreading awareness, these games teach you how to game the system. Can you fake surprise for the camera? It’s not that difficult, really. And what about a smile? As anyone who has had to feign interest at a family event will tell you, the fake smile is handier than the real one. The limits of the intelligent machine are most apparent in the wink-blink game. Socialised human beings know what a wink can mean — a lewd gesture, the acknowledgement of conspiratorial camaraderie, a hint at mischief. Computers still haven’t gotten a hang of all that context.

Of course, these little flaws won’t stop the surveillance industry. And looking at the state of the battle between freedom and privacy, dystopias seem like prophecies. So, think of the games as a little insurance. After all, with all the machines watching, it’s always best to have a fake smile ready.

Source: Read Full Article