In an earlier post I showed what I thought was an impressive video only to be informed by commenters that it was a CGI fake. So once again I had been fooled by a fake video into thinking it was the real thing. The techniques have become so sophisticated that people can now create ‘deep fakes’, where images of one person are superimposed onto videos of someone else. These are done so seamlessly that it is almost impossible for ordinary people casually watching a video to detect that the person they are seeing did not actually say or do the things that we see with our own eyes.
Apparently the software to do this does not require all that much sophistication to use and thus the potential for malicious actors to exploit it is huge. The video below explains what is going on and how some people are trying to find ways to more quickly identify deep fakes, initially focusing on all the candidates for the next presidential election whom mischief makers are most likely to target. (I think this video is real but what do I know?)
One of the most terrifying applications of this technology is to insert Nicholas Cage into every film.
The lesson to be learned is that if you come across a video that you find noteworthy for any reason, the prudent reaction is to assume it is not true until you investigate and have corroborating evidence. This is going to be a hard habit to acquire since we have got so used to thinking of video as being almost as good as being there. All these things suggest that while the internet has sped up the distribution of text, images, and videos, it has effectively slowed down the spread of actual knowledge since we now have to be skeptical.
But there is one benefit that deep fakes could provide and this is when it comes to dubbing the dialogue in films into another language. I personally am not bothered by subtitles but I know that some do not like them. But in dubbing, there is usually at least a small mismatch between lip movements and what we hear, and that can be really off-putting. With deep faking techniques, it should be possible to change the lip movements to match the spoken words in whatever language they have been translated to.