Timnit Gebru says AI doom talk distracts from current harms

In short: AI researcher Timnit Gebru says talk about AI wiping out humanity is harmful because it pulls attention away from today’s real problems.

What's going on

Timnit Gebru, a well known critic of the AI industry, made her case in a new interview on Wired’s Uncanny Valley podcast. She argued that repeating warnings about “existential risk” from AI, meaning AI causing human extinction, has become a familiar cycle.

Gebru said the same wealthy backers who fund major AI companies also fund groups that warn about extreme future dangers. She claimed this can make the warnings sound independent, even when the money comes from overlapping sources. She also said this kind of messaging can help companies market their products as unusually powerful while pushing regulators to focus on far-off scenarios.

Gebru wants more attention on issues that are already here, like energy use from data centers, workers being paid to label data, and companies not clearly documenting what data they used for training. She also pointed to “deceptive marketing,” and argued that existing laws can be applied to AI companies.

The interview revisited her 2021 “stochastic parrots” paper. The phrase is a metaphor for large language models, which are the text engines behind chatbots like ChatGPT and Claude. Her point is that these systems often sound confident while mainly predicting likely next words (like an autocomplete that got much bigger), which can lead people to trust wrong answers.

What to watch

A key question is whether governments treat AI as something that needs brand-new rules, or whether they enforce current laws around truth in advertising, data rights, and accountability. Gebru also called for more transparency, including clearer documentation of training data, so outside experts can check company claims instead of relying on press releases.

Source: Wired