Timnit Gebru: AI Doom Talk 'Meant to Distract' from Real Harms

Gebru's argument did not emerge in a vacuum. She was hired by Google in 2018 to evaluate biases in its AI tools, and she and her fellow researchers presented a paper on the potential dangers of probabilistic large language models. Google rejected it, and she wasn't at the company much longer. That paper—the 'stochastic parrot' paper written in 2021—argued that OpenAI had hyped GPT-2 as too powerful to release and that the industry was asking the wrong questions. Her position today is a continuation of that same critique.

The article places her comments against two incidents: the fight over a 'million-dollar math problem' and whether OpenAI borrowed other researchers' work to solve it first, and the public resignation of an Anthropic researcher who had previously worked at OpenAI, citing how both companies were handling AI safety. Another Anthropic technical staffer responded that people at the company 'really do earnestly believe AI could kill all humans,' adding: 'I personally think it is >10% within the next decade.' Gebru argues these concerns are 'part of a regulatory capture strategy,' and says the race is 'all just cutthroat right now leading up to their IPOs.'

Gebru's core point appears to be that the framing of AI as a potential extinction-level threat—what she calls the 'machine-god narrative'—serves to divert attention from concrete dangers like autonomous weapons, climate exacerbation, and the use of AI by bosses as 'an excuse to get rid of their pesky workers.' Whether this read gains traction may depend on whether the current wave of high-profile safety resignations and public disputes shifts public and policy attention toward documented near-term harms or remains focused on speculative existential risk.