In 1968, I went to work as an editor at The Film Group in Chicago. We made commercials for money and documentaries for love. I was a green kid from the suburbs. Lucky for me, the cinematographer, Mike Shea, loved me. Mike was friends with an editor named Howard Alk. Howard made films with Bob Dylan and gritty civil rights-anti-war documentaries. He was constantly being followed by the FBI.
Once I ran into him at a hospital while visiting a friend. He was a patient across the hall. I saw holes up and down his arms and exclaimed, “Howard! How many blood tests are they giving you?!” He liked me, too. Go figure.
One day, Howard filmed a chef at the Sara Lee bakery who refused to say what the ad company wanted him to about a product. Howard shamelessly cut dialogue transferred to 16mm magnetic tape into syllables and put them back together to create words the chef never said, lip syncing the fake dialogue to the picture. Sensei!
Fast forward to fake news today. I wish Howard were here to see it. Holy moly! I am not talking about tweaking truth or in-your-face lying. I am talking about technologies that are really, really scary like Voco and Face2Face.
Simply explained, Voco records someone speaking. According to RadioLab’s Simon Adler (https://www.radiolab.org/story/breaking-news/), it’s best to collect 40 minutes of vocal content, because that gives an algorithm a lot to work with. The algorithm analyzes speech, figures out the phonetics and sounds of the words, chops them up and creates a data base of sounds for that particular voice. Then…, here it comes…,you type any text you want, and Voco writes a perfect audio track in that particular voice. You can make people say things they never said in their own voice. You could make Trump and Hillary say they love each other. I could have myself give a speech in Hindi. You could make a General’s voice tell his Captain to make his troops go left instead of right.
There is, also, a visual version of this fakery. Up until now, most facial animation has been done with motion capture. A live actor is wired up to a computer which captures and analyzes facial and body dynamics, then transfers that data to an animated figure.
How about doing that in real time, no wire-ups, to the face of a real person? Grail Lab at the University of Washington has developed a program where they feed videos of faces into computers. According to Era Kamelmaker-Schlitzer, head of the lab, the computers are trained to break down faces into a series of points, 62,500 points on one human face. Tracking a person’s face in real time, the program can apply that source person’s facial movements onto a target person for near perfect puppetry.
Put together, this means anyone’s face can be captured and animated to say anything in the sound of their own voice. If the source person speaks in a different language, the target face would be complete with appropriate facial and cultural expressions. Capture a voice, capture an image, add a dash of algorithms, and, in the words of Simon Adler, we raise fake news to a new level. Imagine how this technology might be weaponized!
At Dartmouth College, Hany Fareed, the father of digital forensics, says, “It’s easier to create a fake than detect a fake…. The pace of the media does not lend itself to forensics.”
We were all told as children, don’t believe everything you read. Now, it is don’t believe everything you see or hear!
What can we believe? John Klein of Tapp Media asks, “How do you have democracy where people can’t trust anything they read, hear or see?” Adler points out how paralyzing this is.
Which is why this column is so important to me. I want to share information about your evolving world and help you understand it, so you can avoid paralysis.
If I were queen of education, I would make Logic and Magic compulsory subjects. Logic to equip people with tools to test the validity of arguments and information. Magic to train people to recognize the art of distraction and how easy it is to fool people. If you are in a charter school—suggest it! It is vital that our kids sharpen their critical thinking skills. Magic lessons? A great birthday present.
Adler thinks the solution to this problem will come from the teens of tomorrow. With the inquisitive spirit of early hackers, some bright kid will develop a code to recognize fake news. Probably my nephew, Rama.
Meanwhile, live with a raised eyebrow. As President Reagan said, “Trust but verify.”
Check it out: http://futureoffakenews.com/videos.html.
Vamos a ver!