"If there was an easy consensus answer, we'd have done it by now, so I don't think anyone knows what to do."
That was Sam Altman speaking at BlackRock's Infrastructure Summit in Washington, DC on March 11, addressing what might be the most important question in AI right now: what happens to workers when the machines get good enough?
The OpenAI CEO has been unusually candid in recent weeks. In February, he called out "AI washing" - companies blaming layoffs on AI when the technology had nothing to do with it. Now he's gone further, acknowledging that the real displacement is coming too, and that the tech industry doesn't have answers ready.
"Managing Abundance" Is Not the Easy Problem It Sounds Like
Altman framed the shift in historical terms: "For centuries, maybe millennia, humans have learned how to structure society to manage scarcity, and now we have to quickly learn the opposite - managing abundance." It sounds utopian on paper. In practice, abundance of machine intelligence means a collapse in the value of human cognitive labor, which is how most people earn a living.
He predicted that by late 2028, the total cognitive capacity running inside data centers will exceed human cognitive capacity outside them. That's roughly two and a half years from now. He also described a near future where AI agents handle multi-day and multi-week tasks independently, performing at the level of a senior employee.
The logical endpoint, which Altman didn't shy away from: a "forever deflationary world" where traditional economic metrics like GDP may stop being useful. A new generation of startups is already building with this in mind. In India, founders are attempting "zero person" companies that run entirely on AI and capital, with no human employees at all.
The Construction Workers Get Hired. The Knowledge Workers Get... What?
There's a telling contradiction in OpenAI's own strategy. The company is building gigawatt-scale data center campuses and has partnered with North American building trades unions to create construction pathways. Physical labor to build AI infrastructure is booming. But the stated goal of that infrastructure is to make intelligence "too cheap to meter" - which directly threatens the knowledge workers who currently earn the most.
Altman said he's "not a long-term jobs doomer" and believes humanity will invent new roles. But he also warned that "the next few years are going to be a painful adjustment" marked by "very intense and uncomfortable debates" over how to reshape society. That's a wide gap between the optimistic long view and the bleak medium-term forecast.
The Honesty Is Welcome. The Absence of a Plan Is Not.
Other AI leaders have made similar noises. Anthropic's Dario Amodei has compared AI cluster brainpower to human capacity. Google DeepMind's Demis Hassabis has predicted a post-disruption "renaissance." But Altman's version is notable for its lack of reassurance. He didn't pitch universal basic income. He didn't promise a smooth transition. He said nobody knows what to do, and that's probably the most honest thing any AI CEO has said on stage this year.
The problem is that OpenAI is simultaneously building the thing that causes the disruption and admitting it has no plan for the fallout. That's not hypocrisy exactly - Altman would argue the technology is coming regardless. But it does mean that the people most affected by this shift are getting honesty without a safety net, which is only marginally better than getting neither.