Imagine a world in which politicians can wave away embarrassing clips of their behaviour as fake, or convincing videos of their opponents admitting to wrongdoing are spread online before an election.

According to experts, the generative AI technology required to make this a reality has arrived, representing a “supercharged” issue for democracy and public discourse.

Professor Simon Lucey, from the University of Adelaide’s Australian Institute for Machine Learning, said the pace of development in generative AI is something he hasn’t witnessed in 30 years of observing and commentating on technology.

“It’s very difficult, particularly in still images or a first reading, for a viewer to be able to tell the difference between something that is machine-generated or not,” Lucey said.

“It could be another supercharged avenue for misinformation.”

Deep fakes, videos or images of someone who has had their face, body or voice digitally altered to appear as someone else, have previously been made by skilled operators using Photoshop and video editing in the past.

The ease of access to technology, such as OpenAI’s Sora video generation, will soon put that power into the hands of anybody willing to pay.

‘Supercharged’ issue

Australians across the country will go to polls in state and council elections in 2024, but deep fakes and AI-generated content are already affecting overseas elections.

Will Berryman, from the Royal Institution of Australia, said an AI-generated voice of Joe Biden calling voters is a recent example of how deepfakes can influence politics.

“We can go back to the Cambridge Analytica example of social media changing parts of the population’s belief. AI just supercharges that,” he said.

“When you see politics as a game and you give tools to people to enable them to game and manipulate that competition, that’s what we get.”

The personal data of 87 million people – including those never engaged with the Facebook app launched by Cambridge Analytica – was sold to political operatives and campaigns, before being used for targeted advertising for Donald Trump’s 2016 campaign and Brexit.

Rebecca Johnson, a researcher in AI ethics at the University of Sydney, said the issues raised by the Cambridge Analytica-Facebook scandal still haven’t been resolved.

“We are 10 years in the future and we didn’t really spend a lot of time addressing the underlying methods that were impacting our democracy and the propaganda,” she said.

“Education of the public is the only way to get through this.”

Education and research

Most experts agree that while we shouldn’t be scared of technological advancements, the way generative AI is deployed should be discussed and debated.

Johnson said the technology is an extension of the people who provide the training data, develop the programs and choose how they are deployed.

“We’re forcing these machines to make some kind of decision about what they’re going to output and those decisions are reflective of what we’ve put into the technology in the first place,” she said.

“You’ve got a lot of centralised money and power and so when people talk about value alignment for these technologies, you’ve got to ask whose values you’re trying to align to?”

People like Sam Altman, CEO of OpenAI, have a huge influence on the future of generative AI. Photo: Getty

The Australian government has recently formed a working group for artificial intelligence experts to offer insights into the area and Lucey, who is a member of the group, said it is important to realise “the opportunity Australia has”.

“We have this amazing country and we have these amazing Australian values,” he said.

“If we don’t build up sovereign capability in this area, we are not going to be able to reflect those values back into the new generation of Australians.”