Opinion: What does life look like after the end of reality? The just-concluded Indian general election, in which nearly a billion people were eligible to vote, suggests it’s a prospect worth contemplating. There has never been an election in which the contestants have so thoroughly embraced the potential of unreality.
“This is the purpose of Mr Modi: to celebrate our sad lives, our pain, our unemployment, our inflation,” declared Bollywood star Ranveer Singh in a video shared during the campaign along with a message urging people to vote for the Congress Party, chief rival to Narendra Modi’s BJP. The video caused a sensation – first because Singh, who was not known for his political advocacy, had said it – and second because he definitely did not say it. The video was a deepfake – someone had cloned Singh’s voice and used generative AI to have it read a script.
Soon after, another video appeared to show Congress leader Rahul Gandhi announcing his intention to resign the party leadership and leave India. Both were identified as fakes by the Deepfakes Analysis Unit, a service backed by Facebook owner Meta, but a minority of fakes get this treatment.
Divyendra Singh Jadoun, who calls himself the Indian Deepfaker, became the go-to for Western journalists trying to understand the phenomenon. As the country’s best-known AI faker, he said he had been deluged with requests for work he considered unethical – usually involving the attribution of damaging comments. He had also refused an approach to make a real video look fake: specifically, to replace a candidate’s face in a real and embarrassing video with another copy of his face – thus creating deniability. The real video could then be dismissed as an opposition fake.
Jadoun was able to speak so freely because he didn’t need the work. His company, Polymath Synthetic Media Solutions, has a tidy business creating deepfake video and audio for clients who want fake videos of themselves. Commonly, candidates want to be seen addressing voters in languages they don’t speak themselves, or to create avatars that seem to personally converse with whoever is watching them on a phone.
Others sought to bring back the dead. The chief minister of the state of Andhra Pradesh paid Jadoun to create a video endorsement from his revered father, who died in a helicopter crash in 2009. A video circulating in Tamil Nadu state appeared to show a speech by the daughter of a Tamil Tigers leader addressing a crowd in Britain. She had died in an airstrike in 2009, aged 23 – yet here she was 15 years later as if that had never happened.
Some of these fakes are labelled, others can be spotted fairly easily and a growing minority require forensic analysis. But the danger, should they become pervasive, is not so much fake news as an eventual refusal to credit real news. We already live with a sector of society that demands to be told what it already believes, true or not.
Artificial intelligence will benefit society in innumerable ways. It will make money and save lives. Its risks lie less in a future of sentient computer overlords killing us or keeping us as pets, more in its ability to transform particular tasks: to do the once impossible at such speed that we have no way to collectively adjust in time.
This isn’t just about politics and the public sphere – I asked some cybersecurity professionals this year about the potential for scams using AI-generated voices and they grimaced. A recent visitor to Google HQ told me workers there are “freaking out” at the challenge of dealing with AI-generated content in search results.
Human intelligence has carried our species through countless challenges. Adapting to the implications of replicating it might be one of the biggest ones.
