Timnit Gebru Says AI 'Existential Risk' Talk Is a Harmful Distraction
AI researcher Timnit Gebru argues in a new interview that claims of AI posing an existential threat to humanity are a harmful distraction pushed by industry investors who stand to profit. She calls instead for focus on concrete issues like data transparency and labor exploitation.
Prominent AI researcher Timnit Gebru says she does not believe artificial intelligence poses an existential threat to humanity, calling the narrative not just a distraction but actively harmful.
Gebru rose to public attention years ago after clashing with Google over a research paper that highlighted bias in the company's AI systems, arguing large language models essentially parrot their training data. The dispute led to her departure from Google, after which she founded an institute researching technology-driven harms.
Challenging the existential-risk narrative
In the interview, Gebru traces existential-risk rhetoric back to 2013, citing figures like Elon Musk and Peter Thiel as long-time proponents. She points out that organizations warning of AI's supposed civilization-ending potential are often funded by the same billionaires who stand to profit most from AI companies' growth and eventual public offerings.
She argues that framing AI as an all-powerful "machine god" serves multiple purposes: it attracts investors, warns governments that rivals might get the technology first, and distracts regulators from concrete issues already under discussion, such as copyright disputes, data center pollution, and the exploitation of data-labeling workers.
Gebru also notes that AI company leaders, including Sam Altman and Dario Amodei, publicly call for global cooperation on AI risk while resisting specific regulatory measures, such as the EU's AI Act. She characterizes this as a form of regulatory capture.
As a solution, Gebru points to concrete steps outlined by former FTC chair Lina Khan: cracking down on deceptive marketing practices, requiring transparency and documentation of training data sources, and addressing labor exploitation among the workers who label data for AI systems worldwide.
