Concerns about "Deep Fakes"
How technology can trick us
By Kristen Intemann, Professor of Philosophy, Department of History & Philosophy
June 23, 2019
Is this photo of Obama real?
It is in fact real... in this case... and was taken by Pete Souza (the former Chief Official White House Photographer for U.S. Presidents Ronald Reagan and Barack Obama).
President Barack Obama at the U.S. Air Force Academy PHOTO: By Pete Souza
But, advancements in computing, artificial intelligence and machine learning are enabling sophisticated manipulations of images and videos. "Deep fakes" are video manipulations that make it look as though people said or did things they did not. Deep fakes are increasingly difficult to identify and there are concerns that this technology might have harmful consequences.
Defending against deep fakes
Sites such as Twitter and Reddit have banned deep fake videos, but computer generated videos, altered images, and audio manipulations can be very difficult to identify because they are quite realistic. Computer scientists have developed some tools for detecting manipulations in videos, though they are racing to keep pace with the advances in deep fake technologies. And, deep fakes can be rapidly spread through social media before they are identified.
Javin West and Carl Bergstrom from the University of Washington, have created a website: Which Face is Real? where you can test yourself to see if you can identify computer generated faces and distinguish them from real faces. But, as you will see, it is very challenging. Nonetheless, there are often tell-tale signs of fake images and videos, that can help distinguish real images from computer generated ones. Test yourself, and then learn about how to identify computer generated images here. Take the test again, and see how you improve!
In addition, it is important to know where a video or image is coming from and whether it is being sent or posted by a reliable source. Typically, if the content of a video is reliable, you will be able to find the content and news reporting about it on multiple reliable and independent sources.