AI Researchers Warn of Extinction Risks From Superintelligence in New Video Series

Insiders Sound the Alarm on Superintelligence

Current and former researchers from leading artificial intelligence laboratories—including OpenAI, Google DeepMind, and Anthropic—have published a series of video warnings outlining catastrophic threats posed by advanced AI. Curated and launched by Palisade Research, a non-profit that studies AI capabilities and motivations, the interviews are hosted on the website frominside.ai and feature industry practitioners speaking out about the potential for superintelligence to cause human extinction.

Existential Risks and the Control Problem

The project compiles around a dozen interviews that detail researchers' concerns over runaway capabilities, why they continue to work on technology they consider perilous, and whether existential risk warnings are merely industry marketing or hype.

Several prominent researchers offered direct estimates regarding the likelihood of catastrophic outcomes:

• Geoffrey Irving, a former researcher at both OpenAI and Google DeepMind, stated: "The chance of human extinction is about a coin flip, in my view."
• Neel Nanda, a current research scientist at Google DeepMind, placed the likelihood at a similarly urgent level, noting that there is "at least a 10 percent chance that it causes human extinction, and that is ridiculously high."
• Daniel Kokotajilo, a former OpenAI researcher, highlighted humanity's inability to govern systems that surpass human cognition. Superintelligent AI "would basically be god-like powerful," Kokotajilo said. "Unfortunately we don't know how to control them at all. And so probably nobody would control them. This is exactly as dangerous as it sounds and must not be allowed to happen."

The published clips organize answers topically, allowing viewers to see researchers confront foundational questions regarding their motives and the technical hurdles surrounding alignment.

A Fractured Safety Landscape

The testimonials underscore a widening debate over what "AI safety" entails and how the industry should respond to rapid capability jumps. Despite the shared alarm among participants regarding catastrophic risk, the video series reflects a broader lack of consensus across the field: researchers and institutions remain divided on clear definitions of safety, and the industry has yet to establish a cohesive, actionable plan to guarantee control over future frontier models.