Artificial intelligence has created a “superupdate” finger

Artificial intelligence is now everywhere. And let’s absolutely not what we’re used to seeing in movies or books, sometimes their capabilities are truly impressive.

For example, you can remember Google Duplex. I think many of you have heard about the wave of porn, where through artificial intelligence in the video Actresses have tripped up celebrities faces and it was done very simply, and the result sometimes was very convincing.

And if you now sit and giggle with the word “porn”, then imagine what will happen in a few years when AI will learn so so perfectly to frame the face that the results obtained can be used in a fraudulent manner, in more serious crimes, I’m not talking about everyday bases a La “here’s how your husband cheating on you.”

Now that artificial intelligence has learned to deceive fingerprint scanners. Not exactly, let me explain.

Experts at new York University taught AI in such a way that the created “cooperatization”. The point is that modern dactiloscopy when scanning do not use the entire fingerprint, only certain segments of it, since we never put a finger is always silent corners are different. And, despite the fact that each print is unique, all have similar elements. These elements and used an AI when you create a imprint which is suitable instead.

While this probability in real-world scenarios is quite small. In particular, for systems in which the probability of error is 1:1000, “superupdate” approach in 23% of cases, and in the case of systems with a probability of 1:100 in 77% of cases. The point is that modern dactyloscopic accuracy is often an order of magnitude higher. For the same Touch ID Apple says 1:50000.

Only here it is only the first steps. And what will happen in ten years?

Leave a Reply

Your email address will not be published. Required fields are marked *