AI Firms Trigger Philosophy 'Brain Drain' From Academia
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Jonathan Birch at the London School of Economics says AI companies are now the biggest employers of philosophy PhDs, with large salaries and stock options triggering "a huge brain drain" from academia
- Shane Glackin at the University of Exeter found that telling a model to break one rule causes it to break many others, because semantic links deep in its training corpus hold "good-coded" and "bad-coded" things together — exactly the kind of logical analysis philosophers can unpick
- Philosophers at AI firms are tasked with alignment (preventing harmful outputs), cutting hallucinations, tackling model biases, and applying theories of consciousness to assess whether models display sentience
- Aaron Kagan, chair of the American Philosophical Association's Committee for Non-Academic Careers, found a naive keyword count suggests 26.6% of AI job adverts mention ethics, safety, alignment, governance or policy — but after removing boilerplate, only about 5% substantively involve that work
- Mahrad Almotahari at the University of Edinburgh is sceptical that industry-hired philosophers will resolve "the thorniest" questions of machine consciousness, but says they are well positioned to help engineers extract higher-level descriptions of what models represent
- Alan Turing's famous test for machine intelligence was originally published in the philosophy journal Mind, underscoring a long-standing historical overlap between the two fields that AI companies are now exploiting commercially
Why it matters: After removing boilerplate, only about 5% of advertised AI roles substantively involve philosophy-grounded ethics or alignment work — suggesting the much-publicised brain drain is narrower in scope than headlines imply. With serious philosophical work on consciousness, agency, and morality now funded overwhelmingly by industry, Birch warns research risks being shaped by commercial expectations, potentially biasing answers to the most consequential questions about what AI systems are doing and whether they can suffer.
