Associated Press and University of Chicago Harris School of Public Policyup to dateThe rapid development of AI tools is likely to intensify on an unprecedented scale as 2024 approaches, according to a new poll.electionAccording to the poll, almost six in 10 adults (58%) believe that AI tools will increase the spread of false and misleading information in next year's election. In contrast, only 6% believe that AI will reduce the spread of false information, while a third say it will not have much of an impact.
The survey also found that only 30% of American adults have usedArtificial Intelligence ChatbotLess than half (46%) have heard or read something about AI tools, such as those used in job applications or image generators. Despite this, the majority agree that candidates should not use AI.
Source Note: The image is generated by AI, and the image is authorized by Midjourney
When asked whether certain ways 2024 presidential candidates could use AI were good or bad, the vast majority said creating false or misleading media (83%), editing or retouching photos or videos for political ads (66%), tailoring political ads to individual voters (62%), and answering voters’ questions via chatbots (56%) were all bad. Majorities of both Republicans and Democrats said creating fake pictures or videos (85% for Republicans, 90% for Democrats) and answering voters’ questions (56% for Republicans, 63% for Democrats) were all bad.
The bipartisan pessimism about candidates using AI comes after it was already used in the Republican presidential primary. In April, the Republican National Committee released an ad generated entirely by AI that purported to show what the future of the country would look like if President Joe Biden was re-elected. The ad used realistic-looking but fictitious photos showing closed stores, armored military patrols on the streets and panic caused by a surge of immigrants. The ad disclosed in small print that it was generated by AI.
Florida Republican Gov. Ron DeSantis, who also used AI in his campaign, promoted an ad that used AI-generated imagery to make it appear as if former President Donald Trump was hugging Anthony Fauci, the infectious disease expert overseeing the U.S. response to the COVID-19 pandemic.
“I think they should run on their own merits, not on their ability to strike fear into voters,” said Andie Near, a 42-year-old woman in Holland, Michigan, who typically supports Democrats. She has used AI tools to retouch images while working at a museum, but she thinks politicians using such technology to mislead could “deepen and exacerbate the effects that even regular attack ads might have.”
Thomas Besgen, a 21-year-old college student in Connecticut, also disagrees about candidates using deepfake voices or images to make it appear as if they said something they never said. “Morally, that’s wrong,” he said. He favors banning deepfake ads or, if that’s not possible, requiring them to be labeled as being generated by artificial intelligence.
Currently, the Federal Election Commission is considering a petition urging it to regulate AI-generated deep fake ads before the 2024 election.
While Besgen is skeptical about the use of AI in politics, he is positive about its potential in the economy and society. He often uses AI tools such as ChatGPT to help explain historical topics that interest him or to come up with creative ideas. He also uses the image generator for fun, such as imagining what sports stadiums will look like 100 years in the future. He said he generally trusts the information he gets from ChatGPT and may use it to learn more about presidential candidates, which only 5% adults said they are likely to do.
Polls show that Americans are more likely to consult news media (46%), friends and family (29%), and social media (25%) for information about the presidential election than information provided by AI chatbots.
Americans are overwhelmingly skeptical of information provided by AI chatbots. Only 5% say they are very or extremely confident that the information is accurate, while 33% say they are somewhat confident. The majority of adults (61%) say they have little or no confidence in the information.
This is consistent with the view that many AI experts warn against using chatbots to obtain information. The large language models of AI that power chatbots work by repeatedly choosing the most plausible next word in a sentence, which makes them good at mimicking writing style but also prone to making up facts.
Adults affiliated with both major political parties generally support regulation of AI. They have positive views of various ways that technology companies, the federal government, social media companies or news media could impose bans or flag AI-generated content.
About two-thirds favor a government ban on AI-generated content that includes false or misleading imagery in political ads, and about the same number want technology companies to label all AI-generated content on their platforms. Biden already laid out some federal guidelines for AI this week when he signed an executive order guiding the rapidly developing technology. The order requires the industry to develop safety and security standards and directs the Commerce Department to issue guidance on labeling and watermarking AI-generated content.
Americans generally believe that it is a shared responsibility to prevent false or misleading information generated by artificial intelligence in the 2024 presidential election. About 6 in 10 believe that technology companies (63%) bear a lot of responsibility in this regard, but about half also believe that the news media (53%), social media companies (52%) and the federal government (49%) also bear a lot of responsibility.