The dizzying pace of AI development is hard for even tech-world news junkies to keep up with, never mind those outside that bubble. But that doesn’t mean people don’t have strong opinions about where things are going, or thoughtful concerns about how fast we should be heading there.
Since its founding in May of this year (and public launch in August), the AI Policy Institute (AIPI) has conducted regular polling on Americans’ attitudes about AI development. Its newest poll shared exclusively with Digital Future Daily ahead of its release on Thursday shows overwhelming, bipartisan concern about the use of AI on the battlefield in particular, and support for strong government restrictions on the technology.
The public is very on board with direct restrictions on technology today, and a direct slowdown, Daniel Colson, AIPI’s co-founder and executive director, told me today. Elected officials aren’t taking this seriously right now, but as its political salience increases that’s going to be increasingly seriously on the table.
The tech industry might be used to treating people mainly as potential customers, if the dynamic of the social media era is any indication. But as politicians cast a more critical eye on Big Tech, it’s significant what the average person thinks about how things are changing, and what should be done about it — especially with technologies as potentially disruptive as AI.
With that in mind, a few key takeaways from the poll, which surveyed nearly 1300 respondents via web panel from Nov. 20 to 21, with a margin of error of 4.3 points (crosstabs here):
Americans largely support restrictions on AI-generated content, especially when obvious harms are involved.
Colson and his colleagues found 39 percent net support for government requiring artificial intelligence companies to monitor the use of AI for racist content, and a whopping 60 percent net support for Preventing AI from being used for impersonations using the likeness or voice of people in a video, image or sound form without that person’s consent as a policy priority.
By comparison, they found a relatively wan 15 percent net support for a ban of all political advertisement using AI-generated images or voices of real people. A relatively significant partisan difference might be one explanation, as AIPI found that such a ban enjoyed 29 points of net support among Democrats compared to just 11 points among Republicans, who have already deployed them to some effect in the 2024 presidential primary.
Americans are worried about x-risk.
Colson emphasized Americans’ concern over superhuman capabilities in AI, with 56 percent net support in the poll for a policy goal of Preventing AI from quickly reaching superhuman capabilities, and 50 percent net support for Treating AI as an incredibly powerful and dangerous technology. There was also 43 percent net support for Slowing down the increase in AI capability.
You basically won’t be able to find an elected official who’s willing to say that ought to be a policy priority, but there’s a big difference between where they are and where the public is, Colson said.
Americans want international cooperation, including with China — especially when it comes to defense.
Support for global oversight of AI was particularly intense when it comes to its integration into weapons systems. Fifty-eight percent of respondents supported an international agreement regulating the use of artificial intelligence in war, and 59 percent supported an agreement to ban the use of Artificial Intelligence (AI) in drone warfare and the nuclear chain of command.
Americans are also largely supportive of general global agreements to regulate AI, although not by as wide margins: 51 percent supported the introduction of a global watchdog to regulate the use of artificial intelligence, and 41 percent supported the introduction of an international treaty to ban any ‘smarter-than-human’ artificial intelligence (AI).
There’s a lot Americans don’t know, but the more they learn, the more worried they are.
For almost every question AIPI asked that featured the option to respond Don’t know, somewhere between 15-20 percent responded that way. That reflects how little confidence Americans have in their understanding of the technology thus far — but Colson says that as awareness has increased, he’s seen an attendant spike in worry and willingness to regulate.
Awareness of AI has generally been increasing, and also skepticism of AI has been increasing, he said, citing Pew polling over the past few years. As it becomes increasingly important over the next few years [as a policy issue], the general direction of the public is more and more concerned, and more and more skeptical.
