AI researchers warn superintelligence is dangerously real

The artificial intelligence community faces renewed scrutiny as a series of video interviews featuring developers from top tech companies reveals stark warnings about the true dangers of advancing toward superintelligence. The project was collected and launched on the website frominside.ai by Palisade Research, a nonprofit organization dedicated to studying AI capabilities and motivations.

The interview collection features insights from current and former personnel across industry giants including OpenAI, Google DeepMind, and Anthropic. Among the most alarming remarks came from Geoffrey Irving, a former employee of both OpenAI and Google DeepMind, who stated during a new interview that the probability of human extinction resembles roughly a coin flip.

Reinforcing these concerns, Neel Nanda, a research scientist at Google DeepMind, noted there is at least a 10 percent chance that AI could drive humanity to extinction, calling that figure ridiculously high. Meanwhile, former OpenAI researcher Daniel Kokotajlo went further by claiming that superintelligent AI would possess god-like power while mankind currently possesses no effective methods to control them.

"The chance of human extinction is about a coin flip, in my view"

Geoffrey Irving

Stock photo for illustration only, not from the actual event

While the full interviews remain accessible, the platform also categorizes segments addressing whether these warnings constitute mere marketing hype and why researchers continue building technology they consider dangerous. Mary Phuong of Google offered a candid perspective, noting that people should certainly be suspicious of her statements because her lab pays her salary. Furthermore, Neel Nanda explained that if he did not believe his daily work actively reduced existential risks, he would simply resign, adding that major tech companies will not halt development solely because individual researchers quit.