One of AI’s Fiercest Critics Says All the Doom Talk Is ‘Meant to Distract Us’
| Source: Wired AI
Tags: Timnit Gebru, AI safety, Anthropic, OpenAI, AI ethics, regulatory capture, longtermism
Timnit Gebru argues that AI extinction fears promoted by labs like Anthropic are a deliberate distraction from concrete present harms — autonomous weapons, surveillance, and labor exploitation — and that doom narratives serve a regulatory capture agenda benefiting the companies stoking the panic.
Details
Timnit Gebru, former Google AI ethics researcher, is pushing back hard against what she sees as a coordinated narrative from AI labs that frames existential risk as the central concern of AI safety. The Wired interview follows two industry events this week: an OpenAI controversy over credit for a $1M math prize, and a high-profile resignation by an Anthropic safety researcher citing concerns about how both companies handle safety. One Anthropic employee publicly stated there is a greater-than-10% probability AI kills all humans within a decade. Gebru's core argument: by elevating hypothetical extinction scenarios, companies redirect public and regulatory attention away from harms already happening — biased systems, autonomous weapons, worker exploitation, and environmental costs. She connects this to effective altruism and longtermism ideological currents embedded in labs like OpenAI and Anthropic, arguing these ideologies drive a regulatory capture strategy where the companies stoking fear also get to write the safety rules. She also questions benchmark selection — specifically the fixation on math, chess, and programming — arguing these domains were chosen because claiming breakthroughs there lets labs declare 'we solved intelligence.' The OpenAI math controversy illustrates this: narrative framing of a breakthrough matters more than peer review. Gebru's forthcoming book 'Deep Unlearning' is due early next year. For policy and enterprise teams, her critique is practically relevant: it argues current AI safety frameworks are industry-captured and that regulation focused on present-day harms matters more than speculative alignment concerns.