Nvidia is happy to take your money whether you're using open or closed models, says Jensen Huang

Nvidia CEO Jensen Huang stated that the company doesn't care which model architecture prevails – open or closed – as long as its hardware supports the infrastructure. Huang emphasized the importance of closed models, noting that there's a place for them in the world, but the idea that all entities would operate on general intelligence is not logical.

At the company's Forge event in San Francisco, Huang defended open weight models during a defense of open models against protectionist pressures from Anthropic and OpenAI leaders. However, he acknowledged the economic and strategic appeal of AI services that are more affordable and less locked-in, which aligns with Fireworks.ai's business model of offering a platform for companies to train and run models tuned to their specific needs.

Huang's endorsement of open models was in line with Fireworks.ai's CEO, Lin Qiao, who highlighted the need for the AI frontier to remain open. However, Qiao's focus on cost savings wasn't echoed by Huang, who instead emphasized the value of specialized intelligence, built by adapting open frontier models with domain knowledge and tools. He suggested that with this approach, businesses can essentially create artificial superintelligence.

Huang also touched upon the significance of agent harnesses turning AI models into agents and the potential breakthrough in hardware-in-the-loop, specifically in the realm of robotic systems. Given Nvidia's interest in robotics, this could prove advantageous for the company in the future.