AGI claims from Brockman, Huang and Musk rest on a term even Anthropic's chief calls marketing

The word behind these declarations has no fixed meaning. Business Insider describes AGI as a nebulous term, often taken to mean an AI model that can perform a wide variety of tasks as well as or better than a human ^[3]. Ben Goertzel chose the phrase in 2002 as a book title, for a speculative "thinking machine" that could beat a human at chess and also learn to drive a car ^[10]. Brockman framed his own verdict as personal. "For me, personally, I do think we're there," he said ^[2]. Factory chief executive Matan Grinberg told the Sources podcast, "I think AGI is already here" ^[8]. Counting Altman, Musk and Huang, five executives in the reporting have said it ^[4][1].

Alan Chan, a research fellow at the AI policy think tank GovAI, separates two capabilities the label merges. A hiring plan depends on the second one, the ability to learn a new job. Some definitions, he said, conflate AI's ability to do new things with "the ability to actually do things in the world" ^[17]. "We might have a system that's really good at coding, but it's just really bad at learning to do new things," Chan said ^[18]. "I think when people talk about AGI, they mix those two things together" ^[20].

Peter Voss, among the first software engineers to use the term, puts the same gap in operating terms. A human, he said, can learn to work at a call center with one or two days of training, and current AI systems struggle with the same task ^[11]. "Right now, to try and do even something as simple as customer support, you need billions of dollars of engineering," Voss said ^[12]. "You need hundreds of thousands of example conversations, and the results are still very poor" ^[13].

The executives have a fair point about speed. "From ChatGPT to o1 to Astra in 4 years," Huang wrote in his post congratulating OpenAI ^[6]. Four years of releases is a real record of model progress ^[6]. Voss's call-center test asks a separate question: how quickly a system becomes useful in a job it was not built for. A staffing decision is made one role at a time, this quarter.

Incentives account for some of the volume. Business Insider places the claims next to the financial pressure on OpenAI and Anthropic ahead of expected IPOs ^[15]. Emily Bender, a professor at the University of Washington, said the companies propagate an "illusion" that chatbots are more advanced than they actually are ^[22]. "AGI, like AI, is a marketing term," she said ^[16]. Anthropic faces the same pressure, and its chief executive agrees with her ^[14].

The trade-off for this quarter is timing. I think an AGI declaration belongs with the rest of a vendor's pitch, and the planning figure belongs to what a pilot in a specific role actually costs. A hiring freeze justified by the declaration commits the company before that figure exists. The cost arrives next quarter, at deployment, where Chan said "it just takes so much time and effort to train it to actually be good at that job" ^[19].