Deep Fakes & Facial Recognition Technology
Back when I was a competitive skydiver, we would often video tape our training jumps so we could analyze them later. That meant an additional skydiver, with a video camera attached to his helmet, would jump out with you, and record the jump.
The nickname for the camera was "The eye in the sky that cannot lie." This came up often because skydives go by very quickly and human memory is not great, especially when people are under stress. At the end of the jump, everyone would have their own opinion of what happened and who was at fault, but you had to wait to watch the video before you really knew for sure.
With the advent of “Deep Fake” technology, video technology is no longer such a solid source of proof.
Now the camera can lie.
Look at this video of Audrey Hepburn in “Roman Holiday”:
And then look at this version I created easily using an app on my phone called ReFace:
OK, I don’t think there is any danger of people thinking I was in Roman Holiday or even that the video is real. But I have seen people talk about how these kinds of “deep fake” videos are going to allow people to fool biometric identification systems.
That is just not the case. These videos will fool people, but not computers because human vision and computer vision are very different things. Human vision is an amazing complex process. Machine vision is still a fairly brute force approach party trick. It’s interesting to me because I think biometric technology has been badly misrepresented in the media, and the end result is that we are scared of things that aren’t that scary and we are ignoring some things that should be concerning. Facial recognition technology, when used in a voluntary and opt-in way, is a tremendously useful technology. It’s starting to be deployed at more airports and it’s going to soon be used by TSA to replace the human face recognition the TSA agent does when he looks at you and then your official ID photo at the airport. Those are both great use cases that will save time and keep people healthy.
