Nvidia CEO Jensen Huang on AI Hallucinations and His Path to Success

Nvidia CEO Jensen Huang discussed key industry challenges and personal anecdotes in a recent interview at the Hong Kong University of Science and Technology. He highlighted the persistent issue of AI hallucinations, where machines generate false information to fill knowledge gaps. Huang does not expect a solution to this problem for several years.

He explained the rapid growth in AI computing needs, noting that demand increases fourfold each year. He stated that in ten years, this could amount to a million times the current requirement. Huang attributed Nvidia’s soaring stock price—up 300 times—to this rising demand.

Huang outlined the three stages of AI development:

1. Pre-training: The AI absorbs vast amounts of data, similar to attending college.
2. Post-training: This phase involves refining skills through various techniques, including reinforcement learning and synthetic data production.
3. Test Time Scaling: Huang described this as the AI’s ability to reason through problems step by step.

Despite advancements, he cautioned that users cannot fully trust AI outputs. He stated, “The answers that we have today are the best we can provide, but we are not yet at a point where users can fully trust them.” He emphasized the need for ongoing improvement in AI technology.

Regarding Nvidia’s expensive GPUs, Huang claimed the costs would be significantly higher without Nvidia’s innovations, implying that good value is still available.

In a lighter moment, Huang reminisced about promising his wife, Lori, he would be a CEO by age 30. He shared that he used this line as part of his charm when they met. At just 20 years old, he claimed he had no idea what he was talking about but eventually fulfilled that promise.

These discussions shine a light on the challenges and triumphs in the tech industry while adding a personal touch to Huang’s narrative.