When President Donald Trump addressed the United Nations in late September, he said the word “artificial” in “artificial intelligence” made the technology sound fake. Instead, he said, his government would start calling it “superintelligence,” or “SI.”
Not long after those remarks, the president made it official. He signed an executive order that instructed federal agencies to use his chosen phrase instead of AI in official communications. What Trump says goes: The Department of Justice has since told its staff to follow suit, and other government departments are likely to follow—not least because the president has now branded anyone calling it AI “THE ENEMY.”
And it’s not just the government—the tech sector has responded to Trump’s pronouncement, too. Elon Musk said on October 4 that SpaceXAI would become SpaceXSI.
The rebrand isn’t just Trump being Trump: AI’s capabilities have improved far beyond their previous limits since ChatGPT was first released in November 2022, and are reaching the point of superintelligence—a term that already means something in AI circles. (Superintelligence, in that telling, is the point at which AI systems can systematically outperform humans.)
However, sticking the “superintelligence” label on a technology—when parts of it definitely don’t meet that standard yet—could make things worse by giving people a false sense of confidence and confusing them further. OpenAI CEO Sam Altman said he thought “superintelligence” was “a more accurate term” than the current branding of AI, though he doesn’t think people will find it “less scary.”
An exclusive survey for Fast Company by market research firm Early Studies validates Altman’s perspective. A poll of approximately 1,000 people in the United States—which the firm says is a statistically representative sample of the population based on its methodology—found that only 1 in 9 people think the rebrand will have a positive effect on the U.S. Furthermore, 53% of Americans think people they know would not be comfortable calling the tech “superintelligence.”
Trump kinda has a point
However, Trump’s unilateral action does raise a valid question about the term we currently use, and whether it’s suitable in the first place. AI has always been a bad way to describe a broad range of technologies.
It dates back to a 1955 proposal by math professor John McCarthy and other academics for a Dartmouth College summer workshop held the following year. During the event, disparate researchers working to replicate the brain in silico (in or on a computer) agreed on “artificial intelligence” as a broad term to cover the full scope of their efforts. Seven decades later, it fits better than it did in the 1950s—but is still a highly disputed term.
Tony Thorne, a lexicologist at King’s College London, says AI’s vagueness has helped the industry thrive—and the technology survive. “The term’s weaknesses—vague and imprecise, ambiguous since the first part looks negative and the second impressive or reassuring—are also its strengths,” he notes.
The term’s use over the intervening decades has helped it stick in the minds of companies, governments, researchers, the media, and the public. A YouGov poll found 53% of Americans preferred “AI” as a way to describe the technology we’re currently using, compared with fewer than 10% who preferred Trump’s favored nomenclature.
One reason might be who’s asking for the rebrand. Thorne thinks Trump’s alternative term is “boosterist” and “triumphalist”—two words that fit well with the entrepreneur-turned-politician but that but do nothing to tamp down strongly felt public anxiety about the technology. “Usage, not regulation, decides a word’s take-up and survival,” Thorne says.
A technological misnomer
Still, some of those looking at the tech most closely think “AI” doesn’t work for the current technology, and arguably never did. Meredith Broussard, a professor at New York University’s Arthur L. Carter Journalism Institute and author of Artificial Unintelligence: How Computers Misunderstand the World, objects to the second half of the phrase, not the first.
“It’s extremely misleading to call it ‘intelligence,’” she says. “It makes people think that there is a brain inside the machine, and there absolutely is not.” So a replacement term, Broussard reckons, ought to include the word “machine” and avoid “intelligence” entirely.
Broussard also thinks technologists should stay out of the conversation. “People in tech are uniquely bad at choosing names,” Broussard points out. The proclivity to borrow names from science fiction or beloved popular culture is an obsession of tech bros, but it can also end up carrying over some of those associations onto the tech itself.
Describing software as “intelligent” encourages people to attach human qualities to systems that don’t have them. Using words like “chat,” “bot,” and “agent” make complicated systems sound familiar, Thorne says. “Super” makes it sound like a superhero, adds Broussard—and that becomes worrying.
As for alternative names, Broussard suggests “calculating machine” is accurate, even if tech PR teams might wince at the idea. Thorne would prefer something like “cognitive computing agent,” or “CCA,” but is equally realistic about the prospects of it sticking. Whatever eventually replaces AI, if anything does, Broussard thinks it needs to avoid repeating the same mistake.
“Any new name has to be firmly attached to reality.”