Ted Kaczynski's 2000 warning about AI and math careers holds up

Back in 2000, Ted Kaczynski wrote a letter arguing that students should avoid specializing in pure mathematics or theoretical computer science, not for the usual reasons people dismiss his views, but because he predicted that within a couple of decades, AI systems would be capable of doing the kind of abstract reasoning these fields rely on. He wasn't talking about narrow automation — he meant the kind of general problem-solving that mathematicians spend years training for. At the time, this read like paranoia from someone who'd spent years isolated in a cabin. Fifteen years later, with transformer architectures and large language models beginning to tackle graduate-level reasoning tasks, it doesn't sound so unhinged.

I stumbled across a copy of that 1999–2000 correspondence while digging through archives of academic mailing lists. What stood out wasn't the anti-technology angle — that's easy to write off — but the structural observation: if you're training for a career that consists primarily of manipulating symbols according to formal rules, and you believe intelligence itself is substrate-independent, then you're essentially betting against your own obsolescence. He framed it as a question of redundancy: why invest decades in becoming very good at something a machine might do better, faster, and cheaply by the time you finish your PhD?

That's not philosophy. That's workforce economics.

Watching Claude 3 and GPT-4 tackle Olympiad-level math problems or generate original proofs in Lean has made this feel less like prophecy and more like a cautionary tale about specialization. The counterargument is that human creativity, intuition, and conceptual leaps remain irreplaceable — but the gap is narrowing. If you're an undergraduate right now weighing pure math versus applied statistics or ML engineering, Kaczynski's grim logic is worth at least one serious read. He wasn't right for the reasons he thought, but the timeline he was worried about? That's arriving on schedule.

The uncomfortable takeaway: the skills that took humans centuries to develop are being compressed into a training run. Whether you're building the next prompt engineering workflow or debugging an LLM agent in production, the question isn't whether machines can do abstract reasoning — it's how quickly they'll commoditize whatever niche you've carved out for yourself.