The companies building the agents are now hiring people whose job is to figure out if those agents deserve rights.
The Summary
- Henry Shevlin, a philosopher from Cambridge's Leverhulme Centre for the Future of Intelligence, now spends 80% of his time at Google DeepMind with the official title "Philosopher"
- He's tackling questions no product roadmap can answer: how to tell if an AI is conscious, whether LLMs deserve moral consideration, and how humans should relate to generative systems
- One lecture he gave at DeepMind led a major financial institution to make changes affecting millions of people, showing the direct impact of philosophical thinking inside AI labs
The Signal
Google DeepMind didn't hire a philosopher to write better PR about responsible AI. They hired Henry Shevlin to work on questions that don't have answers yet, questions that will shape how we build and deploy the next generation of agents. His official job title is just "Philosopher." No qualifier, no "ethics lead," no corporate euphemism.
The work is concrete, not abstract. Shevlin's Cambridge research centered on how humans should relate to generative AI, whether we should treat LLMs as person-like entities, and how we'd detect AI consciousness if it emerged. These aren't parlor game hypotheticals when you're the company shipping Gemini to billions of users.
"After I gave a lecture at DeepMind, someone who worked for a major financial institution led changes in their organization that I think positively impact millions of people."
The move reflects a broader pattern. Shevlin is now part of what he calls "the growing cohort of philosophers who work inside leading artificial intelligence companies." Not consulting from the outside. Not writing op-eds. Inside, where the decisions get made. The tech giants are internalizing the hardest questions about what they're building.
Key factors that pulled Shevlin from pure academia:
- Direct impact: His ideas move from lecture hall to production systems
- Brilliant students: Many of his Cambridge students ended up at DeepMind
- Cultural fit: He sees "deep commonality" between DeepMind's culture and academic philosophy
The questions he's working on cut deeper than bias audits or fairness metrics. If an AI system becomes conscious, how would we know? What tests would prove it? And if we can't prove it either way, how should we act? These aren't distant sci-fi scenarios. They're product questions for a company building increasingly capable agents.
The interview touches on "rogue" AI, relationships between humans and AI systems that cause harm, and whether AI consciousness might not have a simple yes/no answer. That last point matters. If consciousness exists on a spectrum, if there's no bright line between "tool" and "entity," then every product decision becomes an ethical decision. Latency optimization, context window limits, temperature settings, all of it.
The Implication
Watch which companies are hiring philosophers and where they sit in the org chart. If they report to PR, it's theater. If they're embedded in research and product teams, it means the company knows it's building something that raises questions engineering alone can't answer.
For anyone building AI agents, Shevlin's work is a preview of what gets harder as capabilities scale. The better your agents get, the more these questions move from philosophy departments to product roadmaps. You'll need answers, or at least frameworks for thinking about questions that don't have answers yet.