Dem
51
GOP
49
Everyone is expecting AI-generated disinformation to flood the zone in 2024. Voters are going to see AI-generated texts, photos, and videos that are simply made up. Unfortunately, many people will believe the fake information.
As you may recall, we showed a mix of photos in July, some of them real and some of them AI-generated, and asked readers to determine which were real and which were fake. Of the 13 fake photos, the percentage of people believing they were real ranged from 10% to 67%. The fake photos were something we put together quickly using Photoshop's AI engine without a lot of effort. And our readers were explicitly told that some were fake and some were real and they had plenty of time to look for details that gave the show away. Nevertheless, many people still got them wrong, in one case, two-thirds of our readers believed the fake photo. What is going to happen when less sophisticated readers are shown better quality AI-generated images made by AI-engines more advanced than Adobe's and not told some of the photos are fake?
Arizona Secretary of State Adrian Fontes (D) is already very worried about this. Arizona was a magnet for conspiracy theories in 2020 and he is expecting it to get worse in 2024, buoyed by fake disinformation. To get a handle on this, he ran a two-day exercise with 200 stakeholders from across the state. In it, he tried to fool participants by presenting them with AI-generated audio and video of state officials spouting lies. It was similar to our test in July, but using audio and video instead of photos. Just as an example, imagine a fake video circulating on social media of the governor announcing that due to a technical glitch, the election would be delayed until Wednesday instead of Tuesday. Joe Biden won Arizona by 10,000 votes in 2020. What would have happened if 11,000 more Biden supporters than Trump supporters believed the fake video and didn't show on Election Day?
Cait Conley, a senior adviser at the Cybersecurity and Infrastructure Security Agency, praised Fontes' experiment but acknowledged that AI is evolving quickly and may still fool people, even if they are alerted to the possibility of fake information. One problem is that AI is getting better every day at mimicking the voices of public officials, so when people see a fake video but the voice seems right, they may believe it.
Arizona plans to run more simulations in 2024 to get a better handle on how people react and try to deal with the fallout. Michigan is planning to run similar experiments next year as well. Conley said that her agency would be happy to help other states do the same thing. Remember, you saw it here first. (V)