This article may seem a little late, but the issue of Deepfakes is still very relevant, and there’s a lot of misinformation spread around about what they are, how they work, and what they’re capable of.
In addition to providing an overview of the matter at hand, we’ll discuss the limits of the technology, and what threat it poses, if any.
In short, Deepfakes are pieces of media wherein another person’s face or body is superimposed onto a picture or video. While photoshop isn’t anything new, Deepfake technology puts a bit of a twist on the situation.
As opposed to more manual methods of photo manipulation that have existed since the start of photography, Deepfakes are far more dynamic and flexible. That’s because of a key distinction in how Deepfake technology works.
Deepfakes employ machine learning techniques. These of course come with pros and cons.
Machine learning is very slow, takes an incredible amount of computing power, and doesn’t result in perfect outcomes every time. For example, the youtuber Tom Scott made a video in 2018 attempting to implement these techniques with his own machine to rather unimpressive results.
The pros; however, are massive. Unlike conventional methods of face swapping, Deepfakes promise to be dynamic and universally applicable. If you train the algorithm well at the start, you can put the desired face onto any target.
This is of course a very scary idea.
Deepfakes are still able to be spotted with current technology and a watchful eye, but it’s getting increasingly difficult. Just last year, the fake Tom Cruise TikTok account took the world by storm, gaining millions of views.
That isn’t to say that you can’t tell at all, but it does raise a question - if somebody wanted to lie using this technology, could they?
Well, not everybody.
It’s still the case that this is very hardware intensive, and isn’t likely to result in a good outcome unless you know what it is that you’re doing. Any amateur with a GPU can not crack open some free software and start cranking out videos of Biden declaring war on China.
That’s not to say that you shouldn't be vigilant and use common sense when analyzing sensational videos of people saying things, just that some of the alarmism of a truly post truth world where it’s impossible to tell what’s real and fake is still a little ways off.