A REAL FACE VS A FAKE AI
People have been taking misleading photos and editing their photos in attempts to be disingenuous before. However, we have never before had to question whether a person in a photo actually exists or not. We have a tendency to trust pictures and it has been this way for a long time. Not any more. We must be aware that AI is now capable of generating pictures of people who don’t and have never existed.
A new website has been created that tests your ability to tell the difference between a real face and a fake one generated by AI. The site is called WhichFaceIsReal.com, created by Jevin West of the Information School and Carl Bergstrom from the biology department at the University of Washington. The website tasks visitors with picking out the real human face when presented alongside a fake one generated by AI.
DEEPFAKES AND THEIR IMPACT
This new website sheds light on a very serious problem in a lighthearted kind of way. However, the problem is quite serious. Aside from generating fake faces deepfakes can portray people doing things they never did in videos or emulate their voice in audio recordings. There is no doubt that famous people and politicians will be the first to feel the effects of such technology.
These deepfakes are also becoming easier and easier to create. At one time you needed sufficient experience with machine learning but now there are free websites that allow you to create your own with little to no effort. In fact, Software engineers from NVIDIA developed the impressive algorithm for generating realistic faces. You may have already heard of it at ThisPersonDoesNotExist.com. At the moment these deepfakes are not 100% convincing, there still are flaws. With time though, they will become indistinguishable from real photos and videos.
The only solution that comes to mind is creating another AI powered program that can flawlessly detect deepfakes. Eventually, a program such as this could exist on every computer, smart phone, tablet, and any other device which could filter out all the deepfakes. In reality of course this would be extremely difficult to pull off but either way this will have to be addressed.
SHEDDING LIGHT ON A SERIOUS PROBLEM
In a post explaining the purpose of their website West and Burgstrom wrote:
“You can’t synthesize a picture out of nothing, we assume a picture had to be of someone. Sure a scammer could appropriate someone else’s picture, but doing so is a risky strategy in a world with google reverse search and so forth. So we tend to trust pictures. A business profile with a picture obviously belongs to someone. A match on a dating site may turn out to be 10 pounds heavier or 10 years older than when a picture was taken, but if there’s a picture, the person obviously exists.
No longer. New adverserial machine learning algorithms allow people to rapidly generate synthetic ‘photographs’ of people who have never existed.”
The duo behind this new website claim their goal is to make the public aware of just how easily these fake images can be generated and how convincing they can seem. According to them, as of now, people who visit their site are successfully spotting the fakes about 70% of the time which is a good sign that for now these images are not all that dangerous just yet.
Some key things to look out for when attempting to spot a fake is looking out for the background of the photo and examining how such things as glasses and hair are rendered. Those are usually dead giveaways that the person is fake. We encourage everyone to see how often they can spot the real face.