How Drexel University is fighting the war against AI-generated images that could plague 2024 campaigns
PHILADELPHIA (CBS) -- On Thursday, we highlighted the threat of artificial intelligence to "supercharge" misinformation and disinformation in the 2024 election.
Some campaigns, including that of Florida Gov. and former GOP presidential candidate Ron DeSantis, have already been accused of using the technology.
"This is really one of the big information security problems of the 21st century," said Drexel University professor Matthew Stamm.
That's how Stamm sees fake media spreading online. Deepfake videos and manipulated images were used to spread misinformation during the 2020 election, but with rapid advancements in artificial intelligence technology, videos and images are now looking more realistic and can be made more easily.'
"You can also say 'make me an image or a video of X politician engaging in a particular behavior,' and it can make a visually realistic image," said Stamm. "Certainly visually realistic at a casual glance, which is often the way that you interact with images on social media."
Drexel's tool to detect AI online
Stamm and his team at Drexel have now created software that's able to spot these fake and manipulated pieces of media. He walked CBS News Philadelphia through the program.
His software spotted a deepfaked video of Ukrainian President Volodmyr Zelenskyy, allegedly calling on fighters to lay down their arms in the country's war with Russia.
"It's reliably picking up that his face is not real. The traces in this don't match the traces in the rest of the video," said Stamm.
One of the people in this photo was not really there. Can you guess who?
Stamm also showed us a seemingly normal image of a family. But when run through the software, it detected that one woman was artificially added in.
Can you tell who it is? Click and drag the bar over to see how the Drexel software flagged this image.
Stamm said the software detected the woman on the far right wearing white was edited into the photo.
AI has already made it into 2024 campaigns
And Stamm says AI has already made its way into the 2024 race for the White House. Last summer, he says his software was able to identify multiple alleged AI-generated images in a campaign ad for Ron DeSantis, made to show former President Donald Trump embracing Dr. Anthony Fauci.
"To the best of my knowledge, this is the first time a major American political candidate used AI-generated media without being forthright that they were using it in the campaign," said Stamm.
With prevalence of AI, users have to take images and video with a grain of salt
But there is one problem. Stamm's software isn't currently available for the public to use. It's only accessed by the U.S. government. That leaves voters on their own to figure out what's real and what isn't.
Stamm says that will take time, education, and a rethinking of the way we interact with video and images online.
"Before, a picture was kind of the truth. Now a picture is more of a suggestion," Stamm said. "You have to look at it in the context of larger evidence and do a little background to say 'is this really real?'"
If you see something outlandish online, experts say the best thing you can do is to question it.
For instance, if you get a call from someone who claims to be a poll worker, and what they're telling you seems off, you can always hang up and call your polling place back.
Also make sure you're checking trusted sources for the best information.
