AI Researchers Warn of Existential Risk from Superintelligence

The project, hosted on the platform frominside.ai by the non-profit Palisade Research, features testimony from experts at leading labs including Anthropic, Google, and OpenAI. Neel Nanda, a research scientist at Google DeepMind, places the probability of human extinction due to AI at a minimum of 10 percent, a figure he describes as intolerably high.

Former OpenAI researcher Daniel Kokotajilo pushes the argument further, comparing the potential capability of future systems to a god-like power. Kokotajilo insists that humanity currently lacks any reliable mechanism to control such systems, asserting that the danger is literal and must be prevented. While the contributors voice a unified alarm regarding the existential trajectory of the technology, the interviews highlight a lack of consensus on specific solutions or regulatory frameworks. The footage also addresses internal skepticism, debating whether these warnings serve as necessary public alerts or are merely contributing to the broader industry hype cycle.