Timnit Gebru: AI 'Existential Risk' Talk Is a Harmful Distraction
Original: Timnit Gebru Believes There Is No ‘Existential Threat’ From AI
Why This Matters
Gebru's critique challenges the dominant safety narrative shaping billions in AI policy and funding decisions.
AI researcher Timnit Gebru, in a new WIRED interview, argues that AI existential risk rhetoric is dangerous and self-serving — a cycle repeated by figures like Elon Musk and Peter Thiel since 2013. Her new book, Deep Unlearning, ships early next year.
Timnit Gebru, who was pushed out of Google after coauthoring a paper calling large language models 'stochastic parrots,' is now taking aim at a different target: the AI doom camp. In a WIRED interview with Lauren Goode, Gebru argues that existential risk narratives around AI are not just a distraction — they are 'harmful.' She traces the rhetoric back to 2013, when Elon Musk and Peter Thiel began publicly warning that AI could destroy humanity, and says the same talking points resurface every few years without meaningful change. Her comments came partly in response to a recent Anthropic researcher's public resignation post on X, in which the researcher claimed many inside the company believe AI poses a genuine existential threat. Gebru sees that framing as strategic rather than sincere — a way for founders to consolidate power and capital under a veneer of saving humanity. She is also responding to critics, including an Anthropic cofounder, who claim her stochastic parrots research is outdated given how much AI has advanced. Gebru's new book, Deep Unlearning: The Rise of AI and the Radicalization of a Tech Idealist, is expected to ship early next year. Her stance has drawn an unusual coalition of allies: her critique of AI's effective altruism community has aligned her, at least rhetorically, with right-wing groups that have similarly targeted EA.