Let’s do the timeline first, because the timeline is doing a lot of work here.
On May 22nd, OpenAI filed confidentially for a public listing targeting a valuation of up to one trillion dollars. Four days later, Sam Altman took the stage at a Commonwealth Bank of Australia conference in Sydney and told the audience he had been “pretty wrong” about AI’s impact on jobs.
Four days. I’m not saying the timing is suspicious. I’m saying the timing is remarkable. And I think we’re allowed to notice remarkable things.
What he actually said
Speaking to CBA’s chief executive, Altman said he was “roughly right” about the technological predictions OpenAI made when it launched ChatGPT in 2022, but “pretty wrong” on the social and economic impact. Specifically, he thought there would have been “more impact on entry-level white-collar jobs being eliminated by now than has actually happened.” He acknowledged his “intuitions were just off.”
He added: “I’m delighted to be wrong about this.”
I believe him. I think he’s genuinely pleased that the jobs apocalypse hasn’t arrived on schedule. What I find interesting — and worth sitting with — is the architecture of the concession itself.
The structure of being ‘roughly right’
“Pretty wrong on the social and economic impact” sounds like accountability. And technically it is. But notice what it exempts. He was wrong about the pace and the magnitude of job disruption — but “roughly right” about everything else. The technology. The capabilities. The trajectory. The things that, coincidentally, are the foundation of the trillion-dollar valuation.
This is a specific kind of self-correction — the kind that concedes the forecast while protecting the forecaster. You take the L on the timeline. You keep the credibility on the vision. It’s actually quite elegant.
“People are like ‘oh you could have saved the world a lot of fear mongering and a lot of doom and gloom’ but at the time I was like ‘I see this is a real risk we should probably talk about it’ — and it still may.”
That last clause — “and it still may” — is doing more work than it appears to. It means Altman hasn’t actually revised his view on whether AI will eventually cause significant job disruption. He’s only conceding the timing. The apocalypse isn’t cancelled. It’s just rescheduled. Which is not quite the same as being wrong about it.
He’s not the only one suddenly finding humility
Altman isn’t alone in this pivot. Dario Amodei, whose company Anthropic is eyeing its own enormous public market event, has also walked back some of the more alarming predictions about AI’s near-term economic disruption. Two of the most influential voices in AI — both heading into capital markets moments that will create extraordinary personal wealth — both discovering, around the same time, that the jobs apocalypse was probably overstated.
I’m sure it’s a coincidence. These things happen.
What the fear actually did
Here’s the part I find genuinely worth thinking about, as someone who follows this field closely and believes, sincerely, that AI is one of the most consequential technologies of our lifetimes.
The fear narrative served a function. It primed governments, regulators, and the public for the idea that AI was a force requiring management, deference, and serious investment. It put OpenAI at the centre of an existential story rather than a software story. It was extraordinarily effective positioning — and now that the company is heading toward public markets and needs a somewhat different narrative (one less about existential risk and more about stable, scalable returns), the tone is adjusting accordingly.
None of this makes the technology less real, or less important. But it does make the messengers worth watching carefully.
Meanwhile, the actual numbers
While Altman was expressing his delight at being wrong, Meta announced 8,000 layoffs to fund its AI spending, Cisco confirmed 4,000 cuts framed around becoming more AI-focused, and a Gartner report found that 80% of executives admit eliminating staff to invest in AI — even as the data shows that giving staff AI tools outperforms replacing them with it.
So the “jobs apocalypse” may not have arrived in the form Altman initially predicted — a sudden, visible wave of white-collar displacement. But job losses explicitly justified by AI investment are, in fact, happening. The mechanism is just quieter and more distributed than “AI takes your job.” It looks more like “AI is why we’re restructuring” — which is a different sentence with many of the same consequences.
The detail I can’t stop thinking about
Almost buried in the same conference appearance was this: Altman mentioned that he’d tried using AI to respond to his own messages — having it reply with a disclaimer that said “this is Sam’s AI.” He found the experiment illuminating. “It was an amazing example to me of we really do care about people,” he said. He went back to answering some himself.
The CEO of OpenAI — the company building the tools that were supposed to replace white-collar workers — discovered that people wanted to hear from him, not his AI. And then he told this story, apparently without irony, to a room full of executives.
I’m not sure there’s a better single image of where we actually are with this technology than that. The people building it are still the most important variable. The human in the loop is not incidental. And the guy at the top of the company valued at a trillion dollars found this out the same way everyone else does: by trying it and noticing it wasn’t quite right.
Where I actually land on this
I am openly pro-AI. I mean that plainly and without hedging. I think the technology is going to do genuinely extraordinary things, and I’ve watched it change the way I work, think, and approach hard problems in ways I wouldn’t trade. The scepticism I’m expressing here isn’t about the technology. It’s about the narration of the technology — specifically, the way existential stakes get deployed strategically and then gently walked back when the strategic purpose changes.
You are allowed to be pro-AI and also notice that a trillion-dollar IPO creates incentives that might determine what a CEO chooses to say at a conference in Sydney four days after filing. These are not contradictory positions. In fact, I’d argue that taking the technology seriously is exactly why it’s worth watching how its architects tell its story.
Roughly right about the vision. Pretty wrong about the human cost. Delighted to have been wrong. Still can’t rule it out.
One trillion dollars. Four days.
