Google DeepMind hires philosopher for AI ethics

Summary

Philosopher Henry Shevlin, who previously served at the University of Cambridge's Leverhulme Centre for the Future of Intelligence, currently spends 80% of his time as a 'resident philosopher' at Google DeepMind. There, he addresses topics such as AI consciousness, human-AI relationships, and agent autonomy. In an interview with IEEE Spectrum, Shevlin defends the creative capacity of LLMs against 'slop' criticisms. He also argues that daily interaction with the technology is essential for meaningful AI ethics work.

Why it matters

The employment of a full-time philosopher by a leading AI lab signals a structural shift. Ethics is no longer an external add-on but is being integrated into the engineering pipeline. Shevlin claims that academia is 'lazy' and that 'genuinely high-quality philosophy and cognitive science are now happening in labs.' This assertion reframes the relationship between industry and academia, potentially creating a transformation that influences who sets AI norms.

Highlights

• Shevlin cites only one anecdote as direct evidence of impact: a lecture he gave leading to a policy change that affected 'millions of people.' However, he explicitly states that he 'cannot go into details' on this matter.
• He distinguishes between 'content slop' and 'LLM-style language,' arguing that the former has dramatically improved. He describes the latter as a correctable prompt artifact involving words like 'dive' and 'load-bearing.'
• He rejects a general ban on AI agent autonomy. He argues that agents should be deployed to accelerate 'boring tasks' and medical research processes, provided safety measures are maintained.

Critical take

The article reads less like investigative journalism and more like a selected Q&A format that allows the subject to frame their own value proposition. Shevlin's impact claims are deliberately vague. Defending LLM creativity against the 'slop' complaints voiced by ordinary users in 2026 appears to be advocacy dressed up as philosophy. There is no independent verification, dissenting voice, or scrutiny regarding whether a single philosopher can meaningfully constrain the safety decisions of thousands of engineers.