Nobody agrees on what AGI actually is. Not the researchers building it, not the labs funding it, not the academics studying it from the outside. And yet the conversation around it proceeds as if there’s a shared definition somewhere that everyone is working toward, there isn’t.
Karen Hao spent years covering AI for MIT Technology review, then The Atlantic. At some point she started asking a question most journalists weren’t asking : who controls the meaning of all this? Not the technology itself but the narrative around it and that question stuck with me. Specially since I read her book “Empire of AI”.
Because the definition of AGI is not a scientific threshold like the boiling point of water. There’s no international body, no peer-reviewed consensus, no agreed benchmark. OpenAI has its own internal definition, Deepmind has theirs and Anthropic largely avoids the term altogether. Each lab sets its own goalposts, runs its own evaluations, appoints its own committees to decide when the line has been crossed.
Sam Altman could wake up tomorrow and announce AGI is here. There’s no external institution that can formally contradict him.
The obvious response is, well, who else would do it? And that’s genuinely fair, A UN committee on AGI would have taken fifteen years and produced something too vague to be useful. Governments don’t have the technical depth and academic consensus moves too slowly for a field that ships every six months. This concentration of definitional power was probably inevitable, maybe even necessary certainly sadly also.
But inevitable doesn’t mean neutral. Because when a lab declares AGI, that declaration unlocks things such as investment rounds, policy conversations or public perception of risk. OpenAI’s own charter includes a clause saying that reaching AGI changes their obligations to investors. The definition isn’t just philosophical, it’s obviously contractual, it’s financial consequences for specific people in specific rooms.
Most people assume there’s a precise technical threshold that experts are measuring against but there isn’t. And that gap between what people assume and what’s actually happening is where most of the confusion comes from.
The labs aren’t necessarily acting in bad faith. But a handful of private companies are writing the definition of one of the most consequential milestones in history, on their own terms, with limited outside input. Whether that was the only realistic path, probably yes. But knowing something was inevitable isn’t the same as being comfortable with it. Shit man.