Q&A with Timnit Gebru, who says the "existential risk" narrative is a harmful marketing strategy and the "stochastic parrots" metaphor is not outdated

The narrative of AI as an existential threat may be a marketing tactic to secure investment and evade regulation.

AI researcher Timnit Gebru has criticized the prevailing "existential risk" narrative surrounding artificial intelligence, asserting it's a deliberate marketing strategy by major tech companies to benefit from IPOs and avoid regulation. Gebru argues that figures like Elon Musk and Peter Thiel, along with founders of companies such as Anthropic, have consistently promoted this narrative of AI posing a threat to humanity since at least 2013. She points out that the same entities, often funded by billionaires who stand to profit from AI companies, also advocate for AI's potential to solve global issues like climate change and poverty. Gebru believes this "machine-god narrative" serves to attract investment, pressure governments, and distract from more immediate concerns such as pollution, data center energy consumption, and copyright issues. She also noted that these companies lobby against and threaten to leave regions with meaningful regulation, while publicly calling for global cooperation on AI safety, suggesting a strategy of regulatory capture.

Gebru's analysis suggests that the "existential risk" discourse is a calculated maneuver by AI companies to position themselves as essential builders of future technologies, thereby attracting capital and deterring governmental oversight. By framing AI as an all-powerful force capable of both utopia and annihilation, these companies can lobby for specific types of regulation that favor their commercial interests, while simultaneously discouraging scrutiny of more mundane but critical issues like data privacy and environmental impact. This strategy also serves to create a competitive advantage by instilling fear of rivals, particularly those in other countries, thus discouraging open-source development and promoting a centralized control over advanced AI.

Furthermore, Gebru contends that the "stochastic parrots" metaphor, which describes AI models as regurgitating their training data, remains relevant and directly challenges the notion that current AI possesses genuine reasoning or consciousness. Critics who dismiss this metaphor as outdated are essentially being drawn into a strategic narrative that elevates AI's capabilities to a level that necessitates the "existential risk" framing. This framing, she argues, allows companies to abdicate responsibility for AI's current harms by attributing potential future dangers to an uncontrollable, superintelligent entity rather than to the design and deployment choices made by human engineers and corporate entities.