Even the founders of generative AI ChatGPT say they are “super nervous” about how autocrats will use the platform.
This new release of artificial intelligence has the potential to unleash disinformation at a pace and scale that is unprecedented.
We won’t be able to fact-check ourselves of this problem.
In the latest episode of Defending Democracy with Malcom Turnbull, Malcolm Turnbull talks to one of the world’s leading voices on AI, Toby Walsh, Author of Machines Behaving Badly.
Walsh says one of the biggest threats to democracy AI is its ability to create text, video and audio which impersonates people.
“You could build a deep fake of Donald Trump for example, and you could get that Donald Trump to ring up every voter in the United States and have an interactive conversation and persuade them to vote for him,” Walsh explains.
“There’s nothing against that, and it’s entirely within the rules. And indeed, the Republican Party have just put out a campaign ad that was entirely deep fake.”
While this particular campaign advertisement is legitimate, Walsh highlights that it can be, along with other similar deep fake videos made, are “deceptive”.
What if the next video an American voter saw of Donald Trump or another politician running for president, was in fact, an impersonation?
“If I were to authorise a chat bot to essentially personate me, it’s not unauthorised but unless you made it clear that it was a chat bot… it would be misleading,” Turnbull says.
“But the real danger would be if somebody impersonated you without authority, programmed it so said, ‘please impersonate Joe Biden in this conversation so that he says he’s an agent of a satanic conspiracy’.”
“We can’t unsee the things that we’ve seen,” Walsh adds.
Defending Democracy with Malcom Turnbull provides a trusted and insightful guide through one of the most consequential questions of our time: Are we witnessing the decline of western/liberal democracy? Listen to the podcast now on the Listnr app.