Generative artificial intelligence tools ‘threaten democracy, reality’

Amid fake images of Donald Trump’s ‘arrest’ and Joe Biden’s struggles, the latest generative AI tools could influence everything from war to elections to reality itself. See the video.

You can’t believe everything you read. But if you see something with your own eyes or hear something with your own ears – well, that’s the ultimate fact-check. Isn’t it?

Millions of Americans recently saw Donald Trump running from arrest, fighting police, and crying in court. They heard Joe Biden announcing conscription for the war in Ukraine, saying extraterrestrial visitors had arrived, and hurling transphobic insults.

Many believed what they scrolled past on social media – perhaps only for a second, or maybe they still do – because that is how our brains are trained.

But none of these images, videos or audio clips were real.

They were the creation of generative artificial intelligence tools, wielded not by genius coders but by anyone with a credit card and an internet connection.

Watch our video above to see if you can tell the difference.

Republican attack advertisement created with AI generated imagery.

In a world where public trust is already frayed by fake news and alternative facts, this threatens to be the development that tears the rope on our mutual connection to reality.

And according to New York University’s AI expert Gary Marcus: “Democracy itself is threatened”.

“Fundamentally, these new systems are going to be destabilising. They can and will create persuasive lies at a scale humanity has never seen before,” he told Congress last month.

Here’s an example. This reporter logged on to ChatGPT and asked the chatbot to create two rather unlikely speeches for Trump and Biden: the former president accepting responsibility for the January 6 riot, and the current president saying he is too old for the job.

Then, with a $US1 payment for voice cloning platform ElevenLabs and a couple of YouTube clips to train the software, ChatGPT’s words came to life in the voices of both politicians.

The whole process took less than 20 minutes, and if you go online and listen to the end result, you will begin to understand Marcus’s point.

He is not alone in sounding the alarm. Hundreds of the world’s AI industry leaders last week released a remarkable 22-word statement: “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.”

Many of them, however, have already unleashed their innovations with limited restrictions and no government regulation.

Another AI fake.

Earlier this year, on the eve of Chicago’s mayoral election, a deepfake video went viral of a candidate defending police officers killing innocent people.

“You can’t stuff the genie back in the bottle once the damage is done,” the Poynter Institute’s Al Tompkins told CBS News. The veteran journalist was explaining the impact of the video on unsuspecting voters, but he also characterised the fallout from the rapid uptake of DIY AI.

Matt Hodges, Biden’s director of engineering on his 2020 presidential campaign, says the depth of the problem is not just voters falling for fakes.

In an interview, he points to Elon Musk’s lawyers recently trying to defend the Tesla owner by challenging the authenticity of a tape of him spruiking the safety of his self-driving cars.

“These tools are able to muddy the water,” Hodges says.

“When we can’t trust the things we see online … then how do we trust the things that are authentic, in addition to avoiding falling for the things that are inauthentic?”

Hodges, who runs political innovation firm Zinc Labs, believes there is a place for AI in politics as a “force multiplier” to draft press releases and fundraising emails, create campaign materials, and produce data analysis and modelling.

‘You can’t stuff the genie back in the bottle once the damage is done.’

When Biden announced his re-election bid in April, the Republican National Committee released an AI-generated advertisement with a dystopian vision of his second term. Hodges says it generated more attention for the method behind the video than its content.

But he warns operatives could be “much more manipulative” ahead of next year’s election, and that as it stands now, there is no “magic tool to tell people if something is fake or not”.

Last month, Biden convened talks at the White House with AI industry chiefs to remind them of their “fundamental responsibility to make sure their products are safe and secure before they are deployed or made public”.

OpenAI chief Sam Altman, the man behind ChatGPT, later told Congress that elections were “one of my areas of greatest concern”. He expressed hope that “the entire industry and government can work together quickly” on appropriate protections.

They are in a race against time. ElevenLabs launched in January, promising “indistinguishable speech quality”, and social media is already flooded with ridiculous but realistic clips of Biden, Trump and others swearing, playing video games and discussing TV shows.

Similarly, images of Trump created from a few lines of text with Midjourney have also gone viral, and while some are clearly fake, University of California computer science expert James O’Brien told USA Today: “I’m very confident in saying that in the long run, it will be impossible to tell the difference between a generated image and a real one.”

The AI era is here – and it’s ready to outsmart us.

Originally published as Generative artificial intelligence tools ‘threaten democracy, reality’