Remember when Barack Obama called President Trump a “total and complete dipshit”? No, me neither. But if you search for it, there’s a video on YouTube where he appears to be saying just that. I used to think that “deepfakes” would change the world. These clips, created using off-the-shelf AI systems, take advantage of new capabilities to near-perfectly edit video, swapping faces, altering expressions and synthesising speech without any artistic expertise.

The technology burst into the public consciousness around Christmas 2016, when a small batch of perverts cottoned on to the possibilities of using it to insert their favourite celebrities into pornography. But others were quick to see the wider ramifications of the tech: a well-timed deepfake of, say, a world leader declaring war, or an FTSE100 chief executive openly discussing their company’s impending bankruptcy, could send shockwaves through the world’s media.

Worse still, if such behaviour became commonplace, it could lead to a breakdown in belief that even extended to real footage. If anything could be fake, everything could be. Seeing is believing, after all, and if you can’t believe what your lying eyes tell you, then what can you believe? Deepfakes would, I believed, usher in an infopocalypse: a new world where commonly held reality fell apart, and chaos reigned.

But then something interesting happened – or rather, didn’t. The year continued, and global information warfare showed no sign of abating, but deepfakes remained curiously absent. Eventually, I realised why: deepfakes are nothing new. We have had the technology to create falsehoods that are indistinguishable from the truth for centuries. Watch: “In a dramatic moment at the close of Monday’s Commons debate, as the government formally deferred the deal vote, the Conservative MP Jacob Rees-Mogg marched forward to grab the mace in protest and held it aloft. The ceremonial object represents the Queen’s authority in parliament – without it parliament cannot meet or pass laws.”

What I did here bears breaking down: using a common piece of software called a “word processor”, I performed a “nameswap” on a piece of text describing events that happened in the House of Commons in early December, substituting the words “Jacob Rees-Mogg” for “Lloyd Russell-Moyle”, and created an alternative reality. I call this a “shallowfake”. You may know it by its older name, a “lie”.

For the vast majority of humanity’s history, information has been almost exclusively transferred through spoken and written language – media that make creating utterly plausible falsehoods easy. One way of looking at the history of journalism is as a gradual evolution of a set of norms and practices to get around that basic vulnerability.

In 1835, the blunt economics of newspaper journalism were such that the New York Sun was motivated to simply fabricate stories of the discovery of life on the moon, falsely attributed to a British astronomer in South Africa. The hoax, published over the course of six weeks, created a media storm. Even as rivals attacked the paper, they also attempted to cash in on the furore themselves; the Sun never formally retracted the hoax, its circulation remained higher for years afterwards.

One hundred and fifty years later, the “Hitler diaries” affair showed how the modern media ecosystem reacts to the same sort of fakery. When an antiques dealer presented the Sunday Times with a “discovery” of 62 volumes of writings by the Nazi dictator, they ran the story as a “world exclusive”. But the tale quickly fell apart – among other red flags, the diaries were written on paper that hadn’t even been available in Hitler’s era – and the Sunday Times was forced into a humiliating climbdown. Rupert Murdoch moved the editor at the time to emeritus position, telling him, according to Charles Hamilton’s book on the affair, “It’s Latin, Frank; the e means you’re out and the meritus means you deserved it.”

The risk of deepfakes, then, isn’t that the newfound ability to create convincing hoaxes may lead to a breakdown in trust and ultimately a debilitating infopocalypse. It’s that the infopocalypse has already happened, and our collective ability to resist fakes of any sort – text or image – is at a centennial low.

The single greatest effect of the internet has been a flattening of the information ecosystem. That allows sources of information to bypass gatekeepers, niche interests to be directly served by publishers who wouldn’t have been able to scale in the days of print, and individuals to interact with each other, in a many-to-many conversation rather than a one-to-many broadcast. Our ability to handle fakes has been reset. On social media, the public is for the first time exposed to the raw firehose of news, with no ability or desire to perform the work of verification, with incentives for sharing the most sensationalist content.

Faced with a race to keep up with the pace of change and an explosion in the availability of new information sources, hoaxes and untruths have gradually infiltrated the pages of even the most respectable journals; the tweets of Russian trolls have made into news coverage, and conspiracy theories such as QAnon have gained such support that they needed to be covered as semi-serious movements. Deepfakes aren’t dangerous because they’ll change the world. They’re dangerous because the world has already changed, and we’re less ready to tackle their reality distortion than we have been for decades.

Alex Hern is a technology reporter for the Guardian

Leave a Reply