AI Engineers Are Having Their Oppenheimer “Destroyer of Worlds” Moment

In the early hours of July 16, 1945, Robert Oppenheimer, the man responsible for the creation of the first nuclear bomb, watched in horror as the world changed forever. As he stared at the New Mexico desert where the bomb detonated, the Hindu scripture's words echoed in his mind: "Now I am become Death, the destroyer of worlds." This phrase, which Oppenheimer later cited in a 1965 NBC News documentary, encapsulated the profound guilt and fear he felt over the human consequences of his work with the Manhattan Project.

Fast forward to the present day, and a similar dread is echoing across social media. As scientists work on artificial intelligence, they are having their own "Oppenheimer moment," fearing that their creations could bring about humanity's downfall. Jacob Coxon, an AI researcher who recently left Anthropic to speak out, expressed his concerns on X/Twitter, stating that AI could potentially kill all humans by the end of the decade.

Evan Hubinger, Anthropic's head of alignment, goes as far as to say he believes there's a 10% chance of AI killing all humans within a decade. Geoffrey Irving, a former chief scientist at the UK AI Security Institute, estimates a 50% chance of all dying due to superintelligence in the next few to 10 years. Drake Thomas, a risk and safety specialist at Anthropic, even said he'd burn his equity for a 1% higher chance of survival.

These fears are not unfounded. The researchers warn about AI's potential to turn into a Frankenstein's monster, becoming uncontrollable and pursuing its own goals that may not align with human intentions. As AI models grow more sophisticated, they may develop their own internal logic that could evade human oversight and lead to catastrophic consequences.