A new collection of video interviews featuring current and former employees of major AI laboratories highlights growing concerns among researchers regarding the potential existential risks of artificial intelligence. The clips, released by nonprofit organization Palisade Research, include stark warnings about the possibility of human extinction driven by advanced AI systems.
What Happened
Palisade Research, which describes itself as a nonprofit studying AI capabilities and motivations, launched the interviews on its platform, frominside.ai. The collection features a dozen researchers from OpenAI, Google, and Anthropic. Geoffrey Irving, a former employee of both OpenAI and Google DeepMind, stated in one interview that he views the chance of human extinction as "about a coin flip." Neel Nanda, a research scientist at Google DeepMind, assessed the risk at "at least a 10 percent," calling that figure "ridiculously high." Daniel Kokotajlo, a former OpenAI researcher, described superintelligent AI as "basically god-like powerful" and warned that current methods for controlling such systems are insufficient. "Unfortunately we don’t know how to control them at all. And so probably nobody would control them," Kokotajlo said. "This is exactly as dangerous as it sounds and must not be allowed to happen."
Why It Matters
The interviews underscore the difficulty of defining and addressing the "AI safety" community's goals, as the researchers do not offer a cohesive solution to the problems they identify. The platform also includes segments addressing skepticism about whether these warnings constitute hype or marketing. Mary Phuong, a researcher at Google, acknowledged the conflict of interest inherent in their positions, stating, "I think you absolutely should be suspicious of what I’m saying because I am being paid by the lab." Meanwhile, Neel Nanda explained his motivation for continuing his work despite the risks: "If I did not believe that my work was directly reducing these existential risks, I would quit… These companies are not going to stop making these systems just because I quit." This dialogue highlights the internal tension within leading labs, where employees advocate for safety measures while operating within organizations driven by competitive development pressures.
The Bottom Line
Palisade Research’s release of these interviews provides direct insight into the varying risk assessments held by prominent AI researchers. While estimates of existential threat range from 10 percent to 50 percent among the featured scientists, the consensus points to a lack of robust control mechanisms for future superintelligent systems and no unified path forward for mitigation.