It makes us all cringe with all our balmy recent phrases. “Rizz” lost its luster when grandparents started asking about its meaning. Teachers who dressed up as “6-7” on Halloween put the nail in the coffin of the battle cry of the alpha generation. And technology CEOs who once trumpeted the quest for “artificial general intelligence” (AGI) are eager to exploit whatever other term they can find.
Until recently, AGI was the ultimate goal of the AI industry. According to reports, it was a vaguely defined term invented in 1997 by Marek Gubrud, a researcher who he defined it as “artificial intelligence systems that rival or exceed the human brain in complexity and speed.” The term still typically means artificial intelligence that equals or exceeds human intelligence. But now several of the largest companies are deciding to rebrand themselves – creating their own phrases or acronyms that (spoiler alert) still mean essentially the same thing.
CEOs have spent the last year downplaying the importance of “AGI” as a milestone. Dario Amodei, CEO of Amazon-backed Anthropic, has he said publicly that he “doesn’t like it”.[s] term AGI” and that he “always thought of it as a marketing term.” – said OpenAI CEO Sam Altman in August that “it’s not a very useful term.” Jeff Dean, Google’s chief scientist and head of Gemini, did just that he said he “dealt with[s] to stay away from AGI talks.” Microsoft CEO Satya Nadella has he said “we’ve gotten a little ahead of ourselves with all the AGI hype” and that ultimately “claiming any AGI milestone” is “just pointless benchmark hacking.” He also said on a recent earnings call that he doesn’t believe that “the AGI as defined, at least by us in our contract, will ever be achieved any time soon.”
Instead, they stuff in a cornucopia of competing terminology. Meta has “personal superintelligence,” Microsoft has “humanistic superintelligence,” Amazon has “useful general intelligence,” and Anthropic has “powerful artificial intelligence.” That’s a keen turnaround for all companies that have bought into the AGI benchmark in recent years and the fear of missing out from failing to meet it.
Part of the problem with “AGI” is that the more advanced AI becomes, the less defined the term seems to be – because the concept of “equal human intelligence” AI looks different for virtually everyone. “Many people have very different definitions of this problem, and the difficulty of the problem depends on trillions of factors,” Dean said.
But some companies are making billions of dollars off this nebulous statement, and the problem is best demonstrated in the strange, ever-changing relationship between Microsoft and OpenAI.
In 2019, OpenAI and Microsoft famously signed an “AGI clause” agreement. This gave Microsoft the right to exploit OpenAI technology until the latter achieved AGI. However, the contract apparently did not fully specify what this meant. When the contract was extended in October, the situation became even more complicated. The terms have changed to say that “when OpenAI declares AGI, that declaration will now be verified by an independent panel of experts” – meaning that now it won’t just be a call for OpenAI to define what AGI means, but it will be a group of industry experts – and Microsoft won’t lose all of its rights to the technology when that happens either. The easiest way to put off this whole ordeal? Just don’t tell AGI.
Another problem is that AGI has acquired some baggage. Tech companies have spent years detailing their concerns about how the technology works could destroy All. Books have been written (think: If anyone builds it, everyone will die). Hunger strikes made headlines. It continued to have a good reputation for a while – claiming that your technology is so powerful that you worry about its impact on Earth seems to attract gigantic investor money. But public opinion, not surprisingly, bitter at the idea. So with complicated definitions, contract dramas, and public fear around super-powerful AI, it’s much easier to promote less loaded terminology. That’s why it seems like every tech company is creating its own brand of “intelligence.”
One popular general-purpose AGI replacement is “artificial superintelligence” (ASI). ASI is artificial intelligence that surpasses human intelligence in virtually every area – compared to AGI, now commonly defined as artificial intelligence that rivals human intelligence. But for some in the tech industry, even the idea of “superintelligence” has become amorphous and equated with AGI. Many theoretical milestones do not even have clearly defined time frames. Amodea says expects “powerful artificial intelligence” to emerge “as early as 2026.” Altman says expects AGI to be developed in the “reasonably near future.”
Therefore, companies have developed their own variants. Meta CEO Mark Zuckerberg said in January that the company must “build for [artificial] general intelligence”, but in July he already did it rotated to “personal superintelligence” in the manifesto. It was a way to give power to people, based on AGI that “helps you achieve your goals, create what you want to see in the world, experience every adventure, be a better friend to the people you care about, and grow to become the person you aspire to be.” Zuckerberg used the manifesto to overcome public fears of AI taking away jobs and to throw shade at Meta’s competitors, calling the company’s vision “different from that of others in the industry who believe that superintelligence should be centrally focused on automating all valuable work, and then humanity will live on the margin of its results.”
However, Microsoft also changed the name of its venture to the pursuit of “Humanistic Superintelligence (HSI)”, which is essentially Zuckerberg’s manifesto in a different font. The company is defining HSI as “extremely advanced artificial intelligence capabilities that always work for and serve humans and humanity more generally” and are “problem-oriented” rather than being an “unconstrained and unconstrained entity with a high degree of autonomy.” The rebranding was completed, among others: new websitetopped with the term “Accessible Intelligence”, supported by a sepia background and a pliable color palette, and filled with images and photos of nature.
From Amazon’s website, yes renamed his AGI’s efforts include the pursuit of “useful general intelligence,” or “artificial intelligence that makes us smarter and gives us greater freedom of action.” Delayed last year, the company hired the founders of Adept, an artificial intelligence startup, and licensed its technology in an attempt to compete with others in the AGI race. However, like other companies’ branding efforts, Amazon is positioning its UGI efforts are useful, straightforward to define, and definitely not all-powerful or scary: they simply “enable practical AI that can actually do things for us and make our customers more productive, empowered, and fulfilled.”
With “powerful artificial intelligence,” Anthropic doesn’t want to seem mundane. Amodei calls the country “a country of data center geniuses” who are “smarter than a Nobel Prize winner in the most important fields – biology, programming, mathematics, engineering, writing, etc.” Powerful artificial intelligence, he said, would be able to write compelling novels, prove unsolved mathematical theorems and write complicated code. It would not only answer questions, but would complete complicated, multi-step tasks over the course of hours, days, or weeks, much like AI CEOs’ vision of an effective AI agent, and “absorb information and generate actions at speeds approximately 10-100 times faster than humans.”
AGI and ASI were already something to be reckoned with. Now we also have PSI, HSI, UGI and PI. Cheers to the recent acronyms that next year will bring.