Eliezer Yudkowsky, a central figure in the effective altruism movement and artificial intelligence research. He has long been a champion for addressing the existential threats that advanced technology can pose. He didn’t graduate high school or college, but he has had a massive positive impact on the AI safety conversation. Consequently, he has become a rockstar in digital elite intellectual circles. His ideas, especially related to the possibility of superintelligent AI, have received a great deal of attention and helped to galvanize public discussion on the future of humanity.
Yudkowsky’s impact largely comes from years of careful thought about the possible risks of AI doomsday. He has spent almost his entire adult life working on this issue. On his personal website, LessWrong.com—which he co-founded—he’s been sounding the alarm about these dangers and trying to foster rigorous, rational discourse. His efforts to raise awareness have positioned him as a key voice in the ongoing conversation about AI’s implications for society.
Yudkowsky joined forces with Nate Soares, the current director of the Machine Intelligence Research Institute. Together, they contributed to and co-authored the book If Anyone Builds It, Everyone Dies. This work examines the unforeseen consequences of developing superintelligent AI, arguing that it could involve technologies we have yet to envision. We have no choice but to focus on reducing the global risk of AI-induced extinction. This is equally urgent as tackling other existential societal dangers such as pandemics and nuclear apocalypse.
Yudkowsky’s predictions about AI being able to create catastrophe are notable for their high level of certainty. He has stated a 99% confidence that superintelligent AI will be an existential threat to humanity. His past statements paint a picture of his go hurry up go approach to these issues. For instance, he predicted that nanotechnology would result in humanity’s destruction “no later than 2010.” So it has gone, critics since deriding his curmudgeonly warnings as “hysterical monomania.” This description does a good job of showcasing the deep seated passions that love this issue.
As Yudkowsky and Soares explain in their book, the most important thing to realize is this. Human cognition cannot match the unpredictable nature of superintelligent systems. They state, “Human thoughts don’t work like that. We wouldn’t struggle to comprehend a sentence that ended without period.” Such a perspective serves to underscore the gulf between how people imagine the capabilities of cutting-edge AI and what it can actually do.
What’s worrisome is that Yudkowsky’s ideas have made their way into the thinking of other prominent figures in the field. Geoffrey Hinton, often referred to as the “godfather of AI,” alongside colleagues Yoshua Bengio and others, echoed Yudkowsky’s concerns by underscoring that “mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.” This impressive, unprecedented agreement from some of the most respected researchers in the field emphasizes the urgent need to address these challenges.
Influential though he may be, Yudkowsky’s approach has drawn a lot of criticism. Notably, Yann LeCun remarked, “People become clinically depressed reading your crap,” reflecting the distressing nature of discussions surrounding AI risks. These reactions show that even though most of us know these conversations are essential, we can—and should—struggle with their uncomfortable implications.
The rapid development of AI technology has been described as “the biggest and fastest rollout of a general purpose technology in history,” according to John Thornhill. This rapid change begs important questions about the ethical implications and regulatory safeguards needed to best address its effects.