On the night of Thursday 3 September 1998, a middle-aged community college professor with a history of heart attacks passed out at the wheel of his car on a busy US highway. The car drifted across the lanes and into the rush of oncoming traffic. The collision was so powerful it thrust the engine of the professor’s car into the front seats. Miraculously, he survived, and no one else was seriously injured. He recovered from a broken ankle and wrist and left hospital. A month later, he was back there with a pain in his leg – a clot that might or might not have been triggered by the accident. Next, his body swelled up to twice its size with fluid, so he looked like a balloon you could prick with a needle and burst. His wife and young children watched as his miraculous survival turned to a sudden worsening of his underlying heart disease. By April 1999 he was dead.
Just over two decades later his son, Michael Grothaus, sat at his computer watching a video of his father, healthy and wearing a yellow T-shirt, playing with a smartphone that was invented many years after his death. He was enjoying himself, recording the sun-dappled park around him. Then he turned towards the screen and smiled benignly at his son from behind his unmistakeable bushy eyebrows.
Grothaus had bought his father back to life as a “deepfake”. It only costs a couple of hundred dollars. There are whole communities of anonymous deepfakers you can easily reach out to in the danker strata of the internet. Usually they specialise in creating made-to-order-porn: say you want a video of yourself making love to Scarlett Johansson, or to the girl next door. All you need to do is provide a video snippet and they do the rest. To create the video of his father in the park, Grothaus sent over 60 seconds of VHS footage from the mid 1990s. “Brad” then broke it down into 1,800 images of his father’s face and ran those images through a program called DeepFaceLab, which grafted them on to a video of another man.
The digital resurrection of his father gave rise to contradictory feelings in Grothaus. He watched the video repeatedly – relishing the reunion. Then he deleted it – horrified at the rupture it had made in reality, and the consequences it implies for our sense of truth and trust.
This split reaction runs through Grothaus’s book on deepfakes. On the one hand they hold out the prospect of overcoming death, envisioning utopia, fulfilling sexual desire. On the other they bring the fear of utter chaos. Even a short fake video of, say, the CEO of a major company resigning, could send markets into panic for just long enough to enable the people who created it to make a killing. Deepfakes of candidates saying something untoward in the final moments of a close election could change the fate of geopolitics.
But while such scenarios are dizzying in their destructive potential, they are, for the most part, still theoretical. The actual financial scam Grothaus describes involves fraudsters who used a voice recording of a CEO to call his accountant and get him to wire them $243,00. Embarrassing – but also only possible because of a pretty gullible interlocutor. The political case study he describes is of an amateur edit of a video that made it look as if Hollywood star Dwayne Johnson was humiliating Hillary Clinton in the run-up to the 2016 election. The video went viral in Magaland, but not because its authenticity was particularly persuasive. It just fitted with people’s existing biases.
That’s the thing about “disinformation”: it’s not really geared towards changing people’s minds. It’s about feeding them what they want to consume anyway. The quality of the deception is not necessarily the crucial factor. Will deepfakes change this? Will their mere existence destroy any vestiges of trust in a shared reality? Potentially. But one thing we do know is that the discourse that has grown up around this issue, rather than being something radically new, is part of a much older dynamic.
Back in a previous life I used to make TV documentaries. I always wondered why anyone agreed to take part in them. Most were ordinary people uninterested in fame. Slowly it dawned on me there was something about the process of filming that seduced them. The camera seemed to promise that their experiences had meaning, and ultimately offered a kind of immortality. That said, whenever our contributors saw the films they featured in, they hated them. The way we edited them into our storylines made them feel less powerful, more vulnerable. Instead of immortality we brought the opposite: a total loss of self-control.
Our relationship with visual representations of ourselves always runs along this axis of narcissism and dread: at once promising a defeat of death, but by arousing that desire only to disappoint it, crushingly reinforcing its inevitability. Our fascination with deepfakes strikes me as the latest iteration of this emotional rollercoaster, and it’s one Grothaus captures very well.