How AI is already changing the 2024 election
Why it matters: There are no rules for using AI in politics. Operatives in both parties are tapping the technology to identify donors and voters more efficiently — and to create photos and videos that reveal the potential risks of "deepfake" messages that could fool voters.
"If 2008 and 2012 were the Facebook elections, this will be the AI election — but it'll be massively more disruptive," said Tom Newhouse, vice president of digital marketing at Convergence Media, a Republican firm that uses AI.
- "I would be willing to bet that the 'October surprise' next year is from AI," he said, referring to the possibility of a game-changing event or revelation just before the November elections.
Zoom in: AI programs are being used for various tasks in just-launched 2024 campaigns, most notably for advertising but also for time-consuming research and office work.
- Immediately after President Biden announced his re-election campaign, the Republican National Committee released an AI-generated video with a series of fictitious images showing a dystopian, chaotic America under a Biden second term.
- The video included scenes from an imaginary Biden and Kamala Harris victory party on election night that contorted their faces and applied ominous shadows throughout.
- Digital firms retained by campaigns are using machine learning to figure out when's best to send text messages to voters, and enlisting AI to quickly identify soundbites from long videos — freeing up staffers and contractors for other work.
- One Democratic pollster told Axios he asked ChatGPT to write a first draft of a gubernatorial candidate's biography, a task that would have taken him about 45 minutes. It took ChatGPT just a few seconds.
Beyond promotional material, AI is being used to improve fundraising efficiency by targeting prospective donors and voters with increasing specificity.
- It allows an upstart campaign with few resources to create just as much content and reach specific voters just as effectively as larger campaigns with more resources.
- In 2020 or 2016, for example, a 35-year-old male might have been targeted as part of a 250,000-person audience, using researchers hired to build databases filled with information from voter files, Newhouse notes.
- Now, a campaign can zero in on that voter as one of 25,000 who may live in a certain Zip code and be a fan of the Washington Nationals or Duke Blue Devils, for example.
How it works: Consultants are using machine learning to sweep large amounts of data — such as a person's location, income and media habits — to find patterns that drive donations or support for a candidate, and create lists ranking likely supporters.
- Democrats who sometimes watch Fox News make great targets for political campaigns, Sterling Data Company, a Democratic firm working with Akkio's AI platform, has found. Sterling figures that's because it takes a certain dedication to Democratic policies to venture to Fox for an opposing view.
- AI is great for "a large candidate — that could be a presidential candidate or anyone who's raising money from a lot more people," says Jonathon Reilly, co-founder and COO of Akkio. "The larger your sorting problem, the more leverage you'll get out of using AI."
Yes, but: AI technology also can be used to flood voters with disinformation.
- It can create AI-generated audio, robocalls, or text messages about a candidate and disperse them to millions in an instant.
- After the RNC ad ran last Tuesday, the White House responded to several queries about whether the material featuring Biden was real or fake.
Thought bubble: The disinformation concern with AI extends beyond fake video images, Axios' chief tech correspondent Ina Fried writes:
- AI can be used to dub over a speech to twist a candidate's words.
- It's a practice that has been used for humor, but incendiary words inserted into an AI-generated video that's made to look as if the candidate actually said them could muddy the waters between fact and fiction.
- And the mere existence of deep-fake technology could be used by politicians to suggest that a piece of genuine but unflattering video is fake.