The pursuit of artificial intelligence that can equal or exceed human cognitive capabilities has long been a primary objective for the tech sector. However, Nvidia's CEO Jensen Huang now views the concept of artificial general intelligence (AGI) as an outdated benchmark.
During an earnings conference on Wednesday, Huang expressed, “We can assert that for many tasks, we've effectively achieved AGI. I feel the discussions around these milestones have become somewhat meaningless.”
In stark contrast, OpenAI’s CEO Sam Altman shared in a recent Time interview that his organization plans to develop AGI by the end of this year, defining it as “highly autonomous systems that surpass human performance in most economically impactful tasks.”
Huang refrained from providing a definition for AGI during the call but emphasized that the landscape of AI is shifting. It's no longer just about users instructing AI to perform specific tasks; instead, a growing number of AI systems, known as agents, are capable of operating independently and enhancing their performance through repetitive self-improvement.
“The key priorities for the industry are that AI is delivering valuable and efficient work, generating profitable tokens, and the potential for increased computational power to produce even greater profits for services,” Huang explained. “This encapsulates our current phase.”
The emphasis on productivity and profitability appears to be a reaction to mounting fears of an AI bubble, especially amid the substantial investments in AI infrastructure. Nevertheless, both OpenAI and Anthropic are yet to demonstrate their profitability. Critics remain skeptical, arguing that large language models might never attain AGI due to their lack of sustained memory, challenges in logical reasoning, and tendencies to produce inaccurate outputs.
Meanwhile, Nvidia is experiencing significant financial success, reporting $96.2 billion in revenue for fiscal Q2, a remarkable 106% increase from the previous year. The company also indicated its growth potential may be limited by an ongoing memory shortage attributed to AI demands, expected to persist at least until early 2028.
Nvidia's CFO Colette Kress noted on the call, “We're witnessing unprecedented price dynamics in memory, with increases surpassing our earlier projections and expected to continue escalating into next year.” This situation poses challenges for Nvidia’s graphics card pricing, which has already risen recently.
Earlier this year, Huang suggested on Lex Fridman's podcast that the tech sector has effectively reached a form of AGI. “I believe it’s already here. We have achieved AGI,” he stated, citing that contemporary AI applications can generate viral content, drawing millions of users despite eventual declines in popularity. “However, the chances of 100,000 of these agents creating Nvidia are absolutely zero,” he concluded.
