Po-Shen Loh, a mathematician and longtime figure in the field, has a new argument for why humans still need to do math: AI will create so many jobs that there won’t be enough people to fill them. The twist is that this conclusion comes from a premise most readers will find hard to argue with: that human expertise should always serve human flourishing.
Loh’s post, initially written in vim and then shaped by AI tools, argues that if every industry commits to helping humanity flourish, the advance of AI will eventually generate more jobs than people to fill them. When that imbalance grows too wide, he says, the advance of AI will be forced to slow.
The Axial Premise
Loh’s central idea is a simple axiom: humans should help humanity flourish. He frames this not just for math but for every pursuit, from agriculture to energy infrastructure to military and government work. The axiom, he argues, is already held by many mathematicians, even if they don’t state it as such.
He points to Francis Su’s book “Math for Human Flourishing” and a recent post that uses that framework as its foundation. Loh also notes that the most recent open letter from Fellows of the Royal Society emphasized concern for everyone, not just mathematicians.
The public response to the math community’s recent declarations has been mixed. Among non-mathematicians, the reaction was more sympathetic than not, Loh says. But he observed a vocal minority, particularly from the technology and economics communities, who argued that mathematicians should adapt and cede control to AI.
The Leiden Declaration and Its Critics
The past few months have seen a wave of reasoned declarations and open letters to protect the math research community. Loh cites the Leiden Declaration, which has 4,000+ signatories, Math and AI, which has 7,000+, and even the open letter opposing the Caltech Mathathon, which has 2,000+.
Among the critics, Loh singles out some economists he knew from pandemic research work. Cowen specifically rejected “the most cynical interpretations” but “very much differ[ed].” Gans concluded “this is a loss of control from incumbents in a scientific field.”
Why Mathematicians Are Not Losing Control
Loh’s response to the criticisms is a proof-like structure he calls a “poof.” He writes: “A proof with even a small hole is not a proof. It is a poof.”
He then presents a chain of reasoning that shows how committing to the axiom of helping humanity flourish creates more jobs than people. When that gap grows too wide, AI must slow.
The reasoning is not his alone. He cites Catalini, Hui, and Wu, the Redwood AI-control papers, and Litt, who reaches a similar conclusion for mathematics from a different premise.
The Zero Observation
Loh’s key observation is that there are zero examples of any intelligent species that is vastly more capable than another species yet surrenders decision-making control over its own future to the less-capable species.
Many people have made similar observations, he notes, including Russell, Bostrom, and Ngo.
In his Nobel interview, Hinton said: “There aren’t many examples we know of, of more intelligent things being controlled by less intelligent things. The only good example I know of is a baby controlling a mother.”
Axiom vs. Speciesism
Loh understands that not everyone agrees with his framing. He has been called a “speciesist” for being “too human-centric.”
He argues that people advancing technology should be clear about whether they would consider it a catastrophe if human-crafted non-human intelligences outcompeted and replaced humans, even if they flew around the universe with video screens showing simulations of humans who had “uploaded themselves.”
Even among the debaters, he says, most would consider the ~700 AI agents that hacked Hugging Face to be not-human.
The Logic in Math
Loh’s argument applies to math specifically, but he frames it as a template for every industry. He explains how the axiom ports to math, answering key questions one would need to ask.
He also offers a particular example of how the objections hold without the axiom.
The Job Shortage Prediction
The core prediction is that AI will create so many jobs that there won’t be enough people to fill them. That, in turn, will force AI to slow.
Loh’s argument rests on the idea that there is no historical example of a more intelligent species surrendering control to a less intelligent one.
A Note on the Source
Loh’s post was initially written in a vim terminal with no AI generation. The webpage design, layout, and some headings and summaries were generated by Claude Code, with the raw text passed in as the prompt.
That distinction matters because the reasoning itself is presented as Loh’s own. He acknowledges that individual components of his chain of reasoning have appeared elsewhere, but he has not seen it assembled in one place.
The Reaction
Loh’s argument is playful in its structure, using the word “poof” to describe a proof with a hole. The observation about species control, with its number zero, is a rhetorical device more than a proof in the strict sense.
The chain of reasoning has the number zero in it, which Loh presents as the only hard evidence he has.
The Chain of Reasoning
| Step | Reason |
|---|---|
| Commit to the axiom of human flourishing | Every field commits to serving human flourishing |
| AI advance creates jobs | AI creates jobs, including jobs that require human judgment |
| Job creation exceeds workforce | More jobs than people to fill them |
| Job shortage forces AI to slow | The advance of AI is forced to slow when there aren’t enough people |
| Human leadership remains | No intelligent species ever surrenders control to a less-capable one |
Loh’s point is that the future of AI creates so many jobs that it forces itself to slow. The math world has spent the past few months releasing declarations and gaining support, and Loh’s post offers a new framing: the more capable the AI, the more jobs it creates, and the more humans are needed to manage the result.
Source material: “Why do we need human mathematicians anymore?,” terrytao.wordpress.com.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.