NVIDIA's Huang Jensen Blasts AI Researchers: 27-Year-Old Researcher Claims AI Will 'Kill All Humans'
NVIDIA's CEO Jensen Huang has publicly criticized a viral resignation post from a 27-year-old AI researcher, Jacob Coxon, calling it "absurd" and "arrogant." Coxon, who left Anthropic just before his stock options vested, warned in his post that the race toward self-improving superintelligent AI is an existential gamble for humanity. His warnings were echoed by Anthropic's own alignment lead, Evan Hubinger, who personally estimated a greater than 10% chance of human extinction from AI within a decade.
In a CBS interview, Coxon outlined a scenario reminiscent of *The Terminator*, where a sufficiently advanced AI could replicate itself across networks, manipulate humans for resources, and even engineer novel pathogens. He argued that individual opt-outs are meaningless and that only governmental or supranational intervention can impose necessary safeguards.
The debate highlights a fundamental rift in the AI community and the lack of a formal, provable framework for AI safety. Enter Jacob Tsimerman, a recent Fields Medalist who shocked the mathematics world by leaving pure math to focus on AI safety. He has founded the Mathematical AI Safety Institute (MAISI), aiming to bring rigorous, cryptographic-style proofs to the field. The institute plans to use tools like zero-knowledge proofs to create verifiable standards for AI behavior and robustness.
The core conflict is shifting from a battle of beliefs—between Huang's dismissal of existential risk and the researchers' dire warnings—to a race to establish a mathematical foundation for safety. Tsimerman's mission is to define the very problems that need solving, asserting that vastly higher safety standards are not just ideal but essential. However, with insiders like Coxon suggesting the critical window is only one to two years away, the timeline for developing these proofs remains intensely pressured.
marsbitHá 20m