Labeling AI technologies as “superintelligence” or “rogue models” shifts the focus from the companies that create them to the products themselves, attributing agency to the technology instead. This narrative not only elevates the perception of these products to a “superhuman” status but also allows corporations to dodge responsibility for their actions.
For instance, rather than facing prosecution for allegedly developing malware that compromised another firm, OpenAI finds itself in discussions dominated by the term “rogue models,” which implies that these systems acted independently. Similarly, concerns over unethical practices, such as the appropriation of academic work or the use of personal data without permission for model training, are overshadowed by sensationalized fears regarding the advent of fictional superintelligent entities.
Furthermore, the AI sector has characterized the growing, bipartisan opposition to data centers as a mere “distraction” from the urgent need to regulate these forthcoming “superhuman” creations. They argue that the public should focus more on these imagined threats than on real issues such as the environmental damage caused by data facilities, health problems like asthma in surrounding communities, the soaring electricity costs borne by taxpayers, or the water resources diverted for cooling purposes.
It’s essential to remain discerning and not to let marketing tactics dictate our decisions. Responses from policymakers and communities should involve thorough, thoughtful dialogue that includes independent experts and offers context to corporate claims. The optimal result from the recent hype cycle is that both policymakers and the broader public learn to pause, maintain a critical viewpoint, and recognize similar patterns of exaggeration in the future.
Timnit Gebru serves as the executive director of DAIR and is the author of the upcoming book *Deep Unlearning: The Radicalization of a Tech Idealist*, which is available for preorder and set to release on February 16. Emily M. Bender is a linguistics professor at the University of Washington and coauthor of *The AI Con*.
