By now you have probably heard about them or maybe you know all about it, but deepfakes are an amazing technology!
They use an AI to continually run passes on an image or to create a face from scratch. Typically, people use this technology to place a famous celebrity’s face on their own or something like that.
Basically, the AI can replace your face and make it look like you were that famous person in the picture or video. This has a lot of benefits because an AI can be used to do all of the hard work instead of animators or artists in a movie production! But there are of course downsides to it as well.
Because deepfakes can make it look like a different person is in a video, this can be used by criminals. Just imagine it for a moment. If someone did a crime then sent a video with your face on their body, it would look like you were doing the crime.
This is a security threat because you can be turned into a criminal without even knowing! But we also see it being done with presidents or other famous people. You can see an example here — Obama Deepfake.
You might have noticed in the video that voices also sound similar to the actual person. While videos may not always use AI for voice, AI has been used to help match voices along with faces. But sound isn’t as hard as faces are, especially when some people naturally sound the same. Either way, whether they use AI for sound or not the deepfakes typically require an AI.
While there is a security risk to deepfakes, AI isn’t perfect. They have to generally use machine learning and you have to provide them with hundreds of images to compare. Sometimes you’ll notice small lip movements that look wrong or maybe other facial gestures.
AI is great but it isn’t perfect! It will probably never be completely identical with face swapping but it can still trick some people.
Another thing to consider is that AI takes time. Because it needs to spend so much time creating the fake image or video replacement it isn’t possible in every use case. In movies we can benefit but when it comes to accuracy it takes a lot of time.
It’s great that the AI can improve the fake facial expressions and replacements but it takes time. The AI might have to run through it thousands of times before the replacement looks realistic and that’s not always possible. However, there are still some great situations to use them!
Deepfakes can help to bring back the dead! Not literally of course, but for movies it can! When characters need to be brought back in sequels or other films we can use deepfakes. Though that does bring morals into question.
This is done by bringing back characters through using their previous film footage. An actor’s face can then be replaced using the deepfake by having the AI match them up. While this isn’t always ideal, it does help add story to some films.
However, this also brings up some moral questions as I mentioned above. Should dead people be brought back to a new film? Is that the right thing to do? I’m not here to make that call, but as long as their families are alright with it and grant the permission then it’s legal. Here is a situation where a family gave permission to use footage of a diseased actor — Carrie Fisher.
That aside, we might also start to see it in the future for actors that can’t make it in. Sometimes an actor gets sick or can’t make it to filming their role. Deepfakes could be used to place their face on an actor’s body if needed.
Overall, this technology is great at showing us another way that AI can be used with machine learning!
With everything above said, to some extent, deepfakes are probably better used to show us what AI can do. Currently the technology itself isn’t perfect and has limited uses. In films it can be a great asset but we must also look at the downside of it.
We can’t use the technology in video games because we are pushing for 60 frames per a second. Video game design would require a lot of time to render out a believable face and it might not make sense to do that.
Another issue is the security risk. It’s dangerous for criminals who use it but famous people would probably avoid using it. For example, if a president had a stand-in it would make more sense just to use someone who looks like them. Likewise, famous people have used similar looking doubles for many years!
Using an AI to create a believable video of you would simply take a lot of time. Sadly, that’s just where the technology is right now. But in the future it could get much faster and at that point we would see it more often!
If anything, deepfakes are a proof of concept. We have seen them prove to us that AI can recognize finer details that we never noticed. It has shown us that just because we see something in a picture or video, doesn’t mean it’s real. It’s a great eye opener to what technology can do and can’t do at the moment.
Because deepfakes are becoming more popular we are seeing them as a security risk. But we are also starting to see potential in them for the future. This could help push the technology further and we might see more realistic renders and uses!
Whether you like deepfakes or you despise them, it has taught us a lot. Machine learning has many different uses and it’s great to see it being used in a different way.
But the key takeaway is that this technology isn’t perfect. It has it’s flaws but we must also remain vigilant to not believe everything we think we see online! Hopefully in the future we’ll see machine learning continue to grow and be used in many other applications.
But until then, what do you think about deepfakes? Should we start to use them more or should we avoid them? Share your comments below!