Warnings from AI Researchers Concerning Superintelligence and Extinction Risk
Current and former employees of leading AI labs such as OpenAI, Google, and Anthropic have published striking interviews suggesting that superintelligence could bring about the end of humanity.
A new video series prepared by Palisade Research has brought to light striking warnings from current and former AI researchers at OpenAI, Google, and Anthropic regarding the existential risks that superintelligence could pose.
Interview Series Launched by Palisade Research
The non-profit AI research organization Palisade Research has made public a series of interviews conducted with current and former researchers in the sector.
Published via frominside.ai, this video series features significant evaluations focusing on the capabilities and motivations of artificial intelligence.
A Coin-Flip Probability for the End of Humanity
Geoffrey Irving, a former OpenAI and Google DeepMind employee, stated in his interview that the probability of human extinction is roughly comparable to a coin flip.
This view lays bare just how serious the concerns within the industry have become regarding the potential dangers of superintelligence.
Neel Nanda's Emphasis on a Ten Percent Risk
Neel Nanda, a research scientist at Google DeepMind, stated that the probability of artificial intelligence bringing about the end of humanity is at least 10 percent.
Nanda underlined that this rate represents an extremely high and alarming level.
God-Like Power and the Control Problem
Daniel Kokotajlo, a former OpenAI researcher, emphasized that systems with superintelligence will attain a god-like power and that they are uncontrollable.
Kokotajlo argued that this situation is as dangerous as it sounds and that it must definitely not be allowed to happen.
Skepticism and the Reasons Behind the Industry's Work
Mary Phuong from Google stated that their own statements should be approached with skepticism because they are funded by the labs.
Neel Nanda, on the other hand, stated that if he did not believe his work directly reduced these risks he would resign, but that even if he quit, companies would continue to build these systems.