NVIDIA CEO Dismisses AI 'Doomsday' Claims, Calls for 80% Resource Shift to Safety

ME AI News, Dongcha Beating AI Bulletin: On the 23rd local time, NVIDIA CEO Jensen Huang delivered a forceful rebuttal in a deep interview with The New York Times to recent warnings from leading labs like OpenAI and Anthropic about AI going out of control. He argued that AI is fundamentally just software and algorithms endowed with objective functions, possessing no “intentions,” and that safety and alignment can be fully addressed through engineering methods such as isolation and sandboxing. “If you believe you’ve lost control, you shouldn’t release the product until you’ve regained it,” he stated, calling it absurd for major tech companies to simultaneously claim their technology is uncontrollable while seeking regulatory or antitrust exemptions. He clarified that his opposition is not to regulation itself, but to using the claim of uncontrollability as a means to evade responsibility. Huang drew an analogy with the chip industry: NVIDIA allocates 80% of its R&D resources to verification, whereas AI labs currently reverse this ratio—spending roughly 80% on capability enhancement and only 20% on safety evaluation, a balance that must be flipped. On computing power and industry outlook, Huang made several startling assertions: future computational demand could increase by a “billion-fold” due to tens of billions of autonomous AI agents invoking large models; NVIDIA’s computing systems are evolving into an asset class akin to commercial aircraft—with a 1-gigawatt AI factory costing $50 billion to build but generating $40–50 billion annually in rental revenue. Huang also revealed a reversal in the open-source vs. closed-source landscape: at the beginning of the year, closed-source models accounted for about 70% of token usage, with open-source at 20%; today, this has flipped to approximately 70% open-source and 30% closed-source, and he emphasized that open-source is the safest cybersecurity strategy. Regarding employment, he refuted claims of job elimination using radiologists as an example, noting that $500 billion in venture capital has flowed into AI-native startups over the past six months, creating numerous new companies. On energy, he proposed a highly disruptive counterintuitive perspective: AI is the “once-in-a-century catalyst” driving global investment in sustainable energy; while short-term increases in fossil fuel use may be unavoidable, this is akin to the temporary pain of surgery—“to save you, you must first feel the pain.” (Source: BlockBeats)