“Welcome to the AGI era,” OpenAI president Greg Brockman declared on a call with reporters last September, touting his company’s latest AI model. “For me, personally, I do think we’re there.” It was a bold proclamation — and, depending on who you ask, either a genuine milestone in the history of technology or a masterclass in corporate hype.
The term AGI, or artificial general intelligence, has migrated rapidly from the fringes of speculative computer science into the daily vocabulary of Silicon Valley’s most powerful executives. Elon Musk has said he “felt the AGI profoundly.” Nvidia CEO Jensen Huang posted a triumphant declaration on X following OpenAI’s release of its Astra model: “AGI has arrived.” Factory CEO Matan Grinberg told a podcast audience that “we are living in a post-AGI world right now.” The proclamations keep coming, and they keep raising the same uncomfortable question: what, exactly, are these people talking about?
A Term Born in a Book Title
To understand why the debate is so fraught, it helps to know where the term came from. In 2002, researcher Ben Goertzel was struggling to find a title for a book about a speculative, highly versatile type of AI — what earlier generations of scientists had called a “thinking machine.” After consulting with colleagues, he landed on artificial general intelligence, a phrase meant to capture the idea of a system capable of performing a wide range of cognitive tasks, not just one narrow function. Think of a machine that could beat a grandmaster at chess in the morning and learn to drive a car by the afternoon.
Peter Voss, a software engineer who contributed a chapter to that book and is widely credited as one of the first people to use the term, describes AGI as a “dream that has really not been realized in AI.” The distinction he draws is crucial. Today’s leading AI systems — the large language models powering ChatGPT, Claude, and their competitors — are extraordinarily capable within specific domains. But they do not learn the way humans do. They cannot adapt fluidly to genuinely new situations without enormous investment in retraining and fine-tuning.
“Humans can learn to work at a call center given one or two days of training,” Voss explained. An AGI system, by his definition, should be able to do the same. Current systems cannot. “Right now, to try and do even something as simple as customer support, you need billions of dollars of engineering,” he said. “You need hundreds of thousands of example conversations, and the results are still very poor.”
The Marketing Dimension
Not everyone in the industry is willing to play along with the triumphalist framing. Anthropic CEO Dario Amodei — whose company is itself a major player in the frontier AI race — has been notably cautious. “AGI has never been a well-defined term, for me,” he said last year. “I’ve always thought of it as a marketing term.”
That candour is worth pausing on. Anthropic and OpenAI are both preparing for what analysts expect to be blockbuster initial public offerings, with valuations potentially reaching into the trillions of dollars. The financial pressure to project an aura of transformative, world-historical progress is immense. Invoking AGI — a concept that conjures images of fully autonomous, free-thinking machines, the kind depicted in films like Her or across the Star Wars universe — serves a clear purpose in that environment.
Emily Bender, a linguistics and AI researcher at the University of Washington, argues that this is precisely the problem. By borrowing language from science fiction, she says, companies propagate an “illusion” that their products are more advanced than they actually are. “We have expectations based on those fictional ideas,” she said. “It refers to an imagined technology rather than any actually existing technology.” Her conclusion is blunt: “AGI, like AI, is a marketing term.”
Alan Chan, an AI research fellow at GovAI, a policy-focused think tank, adds a more technical layer to the critique. The concept of AGI, he argues, conflates two very different capabilities: the ability to perform tasks across a wide range of domains, and the ability to genuinely learn to do new things in the world. “We might have a system that’s really good at coding, but it’s just really bad at learning to do new things,” he said. “So, it’s really good at coding, but when we deploy it in a new job, it just takes so much time and effort to train it to actually be good at that job.” When executives declare AGI has arrived, Chan suggests, they are quietly blurring that distinction.
The Technical Limits of the Current Approach
Beneath the marketing debate lies a genuine and unresolved scientific question: can the current generation of AI systems — built on large language models, trained on vast datasets, powered by ever-larger clusters of specialized chips — actually get us to anything resembling true general intelligence?
A growing number of prominent researchers say the answer is no. Voss argues that the dominance of large language models has crowded out alternative research approaches that might be more promising. “They sucked all of the oxygen out of the air for any other approaches,” he said. Yann LeCun, a foundational figure in modern AI research and Meta’s former chief AI scientist, made a similar point last year. “You cannot just assume that more data and more compute means smarter AI,” he said.
Even within OpenAI itself, doubts have been voiced. Ilya Sutskever, a cofounder of the company, said in 2025 that the industry would need to shift back toward fundamental research. “These models somehow just generalize dramatically worse than people,” he observed. “It’s super obvious. That seems like a very fundamental thing.”
Bender goes further still, arguing that the concept of AGI is not merely distant but categorically impossible. “A general-purpose thinking machine is not within the realm of possibility,” she said. “That would require an approach to engineering that just doesn’t work.”
Why the Words Matter
It would be easy to dismiss this as a semantic squabble among technologists — interesting to specialists, irrelevant to everyone else. But the language that surrounds AI shapes public understanding, policy decisions, and the allocation of enormous resources. When powerful executives declare that AGI has arrived, they are not simply making a technical claim. They are shaping expectations, influencing investors, and nudging regulators toward particular conclusions about how advanced — and how dangerous, or how beneficial — these systems really are.
For Voss, who helped coin the term more than two decades ago, the current moment is a source of genuine frustration. “It’s all just marketing talk, and it’s complete nonsense,” he said. “If you had AGI right now, we would certainly know it.” The machines, for all their impressive outputs, have not yet learned to drive themselves to that conclusion.
