Gebru calls OpenAI and Anthropic doom talk a distraction

The AI industry recently faced turbulent debates following a dispute over a $1 million math problem solved by OpenAI and the high-profile resignation of an Anthropic researcher. In the wake of these events, an Anthropic staff member claimed there is a >10% chance that artificial intelligence could wipe out humanity within the next decade. However, prominent computer scientist Timnit Gebru has dismissed these existential fears, labeling the doomsday narrative as a corporate distraction designed to shield tech giants from accountability.

Gebru, who was hired by Google in 2018 to evaluate algorithmic bias before her highly publicized departure, has long challenged the industry's marketing. She co-authored a landmark 2021 paper on the dangers of large language models, which she famously compared to stochastic parrots. During that era, OpenAI claimed that its GPT-2 model was too dangerous to be released to the public. Gebru argues that such claims, along with the current focus on solving abstract math puzzles, are calculated moves to build hype and inflate valuations ahead of anticipated initial public offerings.

Instead of focusing on hypothetical rogue machines, Gebru urges the industry and policymakers to address tangible, immediate threats. She points to the Leiden Declaration, which cautions against corporate exploitation of the mathematics field to influence policy. According to Gebru, the true existential risks of AI include its integration into autonomous weapons, its massive carbon footprint exacerbating the climate crisis, and its deployment by employers to displace workers. She argues that privileged tech workers obsess over sci-fi scenarios because they are insulated from real-world harms like police brutality or environmental disasters.

For AI practitioners and developers, Gebru's critique suggests a shift in how safety should be integrated into engineering workflows. Rather than designing systems around speculative, long-term alignment theories, engineers must focus on immediate, measurable impacts. This means prioritizing rigorous testing, bias mitigation, and environmental efficiency over hype-driven benchmarks. Practitioners are encouraged to treat AI models like physical infrastructure, ensuring they are built with strict regulatory permits and safety standards rather than relying on corporate press releases.