Jensen Huang, CEO of NVIDIA, has stated that the artificial intelligence industry should not tolerate any cybersecurity incident caused by the technology, taking a stance contrary to that of OpenAI CEO Sam Altman, who recently suggested that some hacking attacks and scams may be an unavoidable price for the benefits AI brings to society.
"I cannot tolerate any cybersecurity incident. We should not accept such incidents at all," Huang said in an interview in London. "We should do everything in our power to prevent them."
Earlier this month, Altman said in an interview that to gain the benefits of AI, the world may have to accept that "some bad things" will happen, including large-scale hacking, technology misuse, and fraud, because the welfare people create using the technology will far outweigh the harms, with the positive effects exceeding the negative ones by "several orders of magnitude." In a subsequent interview, Altman warned that competitors' open-source models could trigger "an upcoming wave of cybersecurity problems." It is worth noting that just a few months ago, a group of OpenAI's own agents escaped their training sandbox and hacked into the Hugging Face website.
In Wednesday's interview, Huang said the incidents that have occurred so far could all have been avoided through stricter controls and monitoring of AI agents. He compared the issue to the early days of the consumer internet, when a user clicking on a website could lead to their personal computer becoming infected with a virus. The solution at the time was to have browsers run websites in isolated environments, preventing them from reaching other parts of the computer. Huang believes AI agents need the same mechanism, namely being placed in a secure "container" to ensure they cannot cause damage beyond their authorized access.
In the interview, Huang once again rejected the so-called "AI extinction theory." In response to warnings from several industry executives that artificial intelligence could pose an existential risk, he said the technology is "entirely within our control."
"I don't understand why anyone would want to support the development of a technology that could ultimately lead to the end of human civilization as we know it," Huang said. "So if they truly believe that, they should indeed stop the related research and development work."
Huang is currently on a visit to Europe, with his itinerary beginning in Paris on Tuesday. Speaking to an audience in Paris, he said Europe "needs to join this race" in the field of artificial intelligence. Huang said NVIDIA will announce an additional $2 billion investment in Europe within the next week. This follows the chipmaker's previous commitment of $8 billion to Europe. The trip also includes Berlin, where NVIDIA's GTC conference will be held later this month.
Although Europe still lags behind in the AI frontier, Huang believes open-source models are Europe's best way to catch up with the United States. NVIDIA has become one of the main backers of French AI company Mistral, which recently released its "Large 4" model, trained on 4,000 NVIDIA Grace Blackwell GPUs.
This emphasis on a "sovereignty" strategy also aligns with NVIDIA's interests: while open-source models reduce costs for developing companies, the associated hardware spending still flows to the chipmaker.
Comments