I have been wanting to write this article for a while now. In fact, I actually wanted to do a video about it, but I just have not found the time. Deepfakes are hitting the web by storm, and the technology that creates them is getting more and more complex, and near indistinguishable from reality. What are the implications of this? What are the benefits? What are the dangers?
So what is a Deepfake? Wikipedia describes Deekfake as:
We have a similar technology which we have used for years where we map one person’s face onto another. One great example of this was Weird Al’s “We Perform this way” where they used the body of an actress and mapped Weird Al’s face in post, however, this is not quite the same as a Deepfake, and in several frames, you can see how the mapping didn’t quite work so well.
Deepfakes are a little different. You do not need tracking marks to superimpose the new image. The Deepfake application simply “studies” the image frame by frame. One of the Deepfake applications has the ability to search the web for images of the person you are looking to superimpose and tries to make the same angles/expressions found in the original movie. Deepfakes also do not need to be full face, they can map just the lips as is the case in this next video of former President Obama. The video is real, all but the lips which have been re-formatted to follow the words of Jordan Peele.
You can tell at some points its fake as the lips do look computer-generated at times, but if you are not paying that much attention, it could pass as real.
Now, most of the Deepfakes that have happened so far are simple and poke fun at celebrities, by superimposing Nicolas Cage or Steve Buscemi. Some of these have been so well done, its eery to look at.
So what are the benefits to this tech? Well, the 9th Star Wars movie is coming out this Christmas and it has a scene (according to the preview) with Princess Leia, however, Carrie Fisher died in 2016. For the movie business, Deepfake Technology could be a great innovation. You no longer need to hire actors to play parts, you could simply license their appearance, and slap it on to any double. Or to a lesser effect, stunt doubles could actually look like the actors they are portraying in stunts.
Then there is the voice aspect, there is an emerging technology that can mimic voice as well, all you need to do is type what you want the person to say:
This technology mixed with Deepfake could also be used for nefarious reasons. And has in the past. The most popular of course has been placing celebrity faces on porn stars.
But it has been said that within 6 to 12 months, it will be almost impossible to tell the fake from the real. This technology could be used to destroy a person’s reputation or instill fear in the population by mimicking a person of power. Video evidence could be forged and no one would be able to tell the difference. The implications of this are quite serious. Here are a few articles I have found on the topic, which go more into detail about the seriousness of this new technology:
- Inside the Pentagon’s race against deepfake videos
- ‘Perfectly real’ deepfakes will arrive in 6 months to a year, technology pioneer Hao Li says
- The rise of deepfakes and what Facebook, Twitter and Google are doing to…
- Deepfake videos ‘double in nine months’