Article
Jensen Huang rebuts AI“doomerism”: Safety is an engineering issue, not a pretext for regulatory exemptions; he calls on the industry to devote 80%resources to safety evaluation
Beating AI News Flash: NVIDIA CEO CEO Jensen Huang, local time 23, pushed back forcefully in an in-depth interview with The New York Times against the recent wave of warnings from frontier labs such as OpenAI and Anthropic that “AI could get out of control.” He believes that AI are essentially software and algorithms assigned objective functions, with no “will” of their own; safety and alignment can be fully addressed through engineering measures such as isolation and sandboxing. “If you believe you can no longer control the situation, then you should not release products until you regain control.” He said bluntly that it is absurd for major technology companies to claim their technology is too powerful to control while seeking regulatory or antitrust exemptions, and clarified that he opposes “seeking liability exemptions on the grounds that the technology cannot be controlled,” rather than regulation itself. Huang drew an analogy with the chip industry: NVIDIA 80%of its R&D resources into validation, while AI labs are currently the exact opposite: approximately 80%is used to improve capabilities and 20%for safety evaluation; this ratio needs to be reversed.
On computing power and industry prospects, Huang made several striking assertions: future computation may increase “a billionfold” compared with the past, as hundreds of billions of AI agents will independently invoke large language models; NVIDIA's computing systems are evolving into an “asset class” similar to commercial aircraft; building a 1-gigawatt AI factory requires 50000000000 USD, but it can be leased out for 400 to 50000000000 USD annually.
Huang also disclosed that the open-source versus closed-source landscape has reversed: at the beginning of the year, closed-source models Token accounted for approximately 70%of usage and open-source models approximately 20%, whereas now open-source accounts for approximately 70%and closed-source approximately 30%. He also emphasized that open source is the safest cybersecurity strategy. On employment, he rebutted job-elimination arguments using radiologists as an example, noting that 500000000000 USD in venture capital has flowed into AI-native companies over the past months, creating a large number of new companies. On energy, he also proposed highly disruptive reverse thinking:AI is precisely the “once-in-a-century catalyst” driving global investment in sustainable energy. Fossil-fuel use may have to increase in the short term, but this is like the short pain of surgery: “To save you, we must first cause you pain.”