For the last week we’ve watched Mark Zuckerburg testify before Congress regarding the role Facebook played to allow fake news to proliferate on its platform during the 2016 presidential election.
We also learned how our personal information was mined and sold to Cambridge Analytica.
We learned that much of the information purchased by Cambridge Analytica was then resold to third parties who then used our information as a blueprint for where to send their various fake news stories. They kept it pretty simple.
Fake news stories that were anti-Trump were sent to individuals who identified with the Democratic Party. Fake news that was anti-Hillary or anti-Obama was sent to those who identified with the right.
The plan was to spark outrage about the Democrats among the Republicans, and vice-versa, to further widen the rift between Democrats and Republicans, and if they were lucky, help sway the election results. It worked – beautifully.
Fake news got widely shared across Facebook and other social media platforms. The vast majority of these fake stories was ridiculously unsophisticated. It was very clearly nothing more than propaganda, even on its face.
And when the origin of the “news” story was tracked back to its source, we found it came from an unknown, certainly unreliable, source.
The next wave of fake news will be even more sophisticated and it will be difficult to determine if it’s true or untrue. It will rely on face-swapping and voice-overlay technology which is in its early stages of development but already quite refined.
Face swapping technology has been around for several years. It’s an algorithm that captures someone’s face and places it over another person’s face, such as putting your friend’s face over Justin Bieber’s.
With past and current face-swapping technology, the result looks crude and is obviously fake. But all that is about to change, and when it arrives, it’s expected to alter the entire media landscape for the worse.
We will see more video, not mere photos. The new algorithms are capable of gathering video snapshots of a person’s face from thousands of different angles and store them in a face library. Then the face is superimposed over an existing face in an existing video frame by frame. It works like this: the existing video is advanced one frame. The algorithm finds a matching angle from the face-library, sizes it to match the face in the original video, and superimposes it.
This new technology is called DeepFake.
According to NPR, the DeepFake technology was created by the porn industry a few years ago with the intent of superimposing faces of celebrities over the faces of an actress in pornographic videos.
NPR reported that Daisy Ridley, Gal Gadot and Taylor Swift have already fallen victim to this new technology.
Samantha Jones, an editor at Motherboard, was one of the first people to report about DeepFake technology.
"Someone takes a dataset of one person's face — and a lot of pictures of that person's face — and then a video that they want to put it on," Cole says. "And they run a machine-learning algorithm, train it on these two images, and after a few hours, gives you the result which is these very realistic, fake porn videos."
What the news industry is expecting next is that DeepFake technology will be used to attack and smear politicians. Politicians, especially those running for president, Senate or Congress, give lots of speeches. It’s very easy to collect a facial library of them and later use these in any fake video they may wish to create.
If a group is creating a fake video of a politician giving a fake speech, it will also need some convincing audio to accompany the video. It’s just as easy to record a library of their words. They’ll use an algorithm to separate those sentences into individual words and then stitch them back together so the fake talking heads say whatever the fake news makers want them to say.
Synching video with voice-over audio used to be tricky. It was easy to spot as fake. But the technology now exists that blends facial movements seamlessly with the rhythm of the audio. In the last few weeks we have seen it employed for what might the first time in history. In the last week or two, many of you have probably seen a video of President Obama giving a speech. The video was 100 percent fake but it looked quite real. The major giveaway that it might be fake is that the voice didn’t seem to be Obama’s. They used Obama’s face and the voice of actor Jordan Peele.
The fake video was created by BuzzFeed. It wasn’t an effort to smear Obama but to show how advanced the DeepFake technology has become. According to BuzzFeed, it took them 56 hours to produce the fake video with the help of a video effects professional.
A producer with BuzzFeed said the technology is still advanced and requires advanced skills to produce a fake video. But that the algorithms get smarter and the technology more user-friendly every day.
What will be most convincing to the average American is a fake-subversive video that seems to show a candidate talking candidly to friends unaware that he or she is being recorded, something similar to the video of presidential candidate Mitt Romney being recorded saying that the poorest 47 percent of Americans were a drag on the nation and its economy.
Imagine if there was fake video of a Democratic candidate who appeared to be caught on tape saying, “To be honest I think programs for poor people are a complete waste of money. Personally, I’d like to see them all get sterilized.” Such a salacious news story would get passed around the internet very quickly. Even if days or weeks later it was shown that the video was a fake, a lot of damage would have been done.
Such videos will keep the candidate on defense and off message as he or she spends several news cycles telling the public that the video is fake.
There was a time when people had morals and ethics and would never consider doing something so blatantly false and damaging to a person’s character. Most people still have principles but enough people and organizations now exist that do not.
It remains to be seen whether the technology will advance quickly enough to affect the 2018 mid-terms. Most news outlets believe it will certainly be ready for the 2020 elections, and that the public will get bombarded with DeepFake videos showing candidates in a bad light.
The technology is here to create sophisticated fake news.
The question is, are humans sophisticated enough to recognize it as fake?