The Death of Meaning
If machines do all the discovering and no human is left who can understand it, does anything actually get discovered? On AI, the career of the mathematician, and where meaning lives.
If a tree falls in the forest, and nobody was around to hear it, did it make a sound? This old saying is a commentary on subjective experience. Sound waves exist, but the conscious experience of them didn’t. What matters?
The cosmologist Brian Cox often reminds us that “humans are temporary structures made of stardust that have the unique capacity to think, feel, and explore the universe”. That sets us apart from rocks banging around, or “normal” chemical and physical interactions. “Meaning” itself is something we experience and hold in our minds. For all we know, we might be the only island of meaning in the entire universe. How precious is that?!
Let’s hold this almost cliché bit of armchair philosophy in mind for a second and switch to the recent drama around the Navier–Stokes Clay Prize (there’s a great overview of it here if you’ve not heard about it).
Update, October 10, 2026: On October 6, OpenAI dropped hundreds of math solutions produced by an internal frontier AI model. In an interview today, Tristan Buckmaster said:
“This recent dump of papers on Tuesday destroyed the careers of early-career mathematicians. We’ve had whole research programs wiped out on Tuesday by their dump … and you don’t even have to directly use an LLM. So if you submit a research proposal, a panelist will look at that research proposal and they have to write a summary. And often how they write the summary is they put it into an LLM … they have all the mathematicians’ research programs in their training data.”
As complex as it already is, I think there is a deeper consequence than just the claims of plagiarism. While it is still unclear if OpenAI’s swarm would have solved the problem without training on the human researchers’ Codex session data — let’s actually assume they didn’t. Or at least assume that one day they do build agent harness swarms capable of doing new math entirely without human mathematicians driving the investigation.
Terence Tao recently published a letter that kinda kept me up last night after reading it. James Maynard restated a similar view in an interview.
I’ll do my best to TL;DR it here.
Mathematics is about advancing the frontier of human understanding. Math just is. Much like the universe exists and just is. The truth is out there whether we discover it… or not. Discovery itself is about exploring this infinite domain of truths and connecting back to our chain of understanding. Mathematicians devote their lives to the quest. Mathematical progress doesn’t really “happen” until that understanding gets absorbed into a critical mass of human minds and connected back to the chain of human understanding.
Vital to this is the ability to build a career as a mathematician. Without credit and attribution, they can’t actually make a living doing this. We can’t decouple the profession from the quest itself. A consequence of AI superintelligence (or AI lab plagiarism) always scooping human mathematicians is that we destroy the career of the mathematician itself. We will never produce new mathematicians.
If AIs produce individual proofs that are hundreds or thousands of pages long of Lean verification code — without building a bridge to human understanding. And there are also no humans left even capable of building that understanding (because we killed mathematics as a career). Does anything actually get discovered?
I don’t know if machines exploring the infinite expanse of the truth mines without connecting humans to new understanding actually creates any meaning… and this really is a fundamental misalignment.
If humans are the source of meaning in the universe. And human discovery and progress — and its associated meaning — is about mapping existence into the collective understanding of humanity. Do we ironically end the meaning of knowledge and scientific progress itself if the machines are the only ones doing the discovery?
There are analogs here to every profession. We were the first to experience this in software engineering, I think. Though in some senses we are luckier than the mathematicians, since software is generally more about the output than the internals. So I know that part is already well explored — and I guess, obvious to many. But it just strikes me that we may destroy the meaning of the pursuit of knowledge itself, while ironically producing the most powerful knowledge acquisition machines conceivable.