In an interview at the Dreamforce conference, NVIDIA CEO Jensen Huang discusses the future of AI as agentic, where AI agents will collaborate with humans and evolve to create their own tools, driven by advancements in unsupervised learning and computing power. He emphasizes the need for effective onboarding processes for these agents and envisions a transformative shift towards dynamic, real-time computing solutions that will revolutionize human-technology interaction.
In an interview at the Dreamforce conference, Jensen Huang, CEO of NVIDIA, discusses the future of artificial intelligence (AI) as being agentic, where thousands or even millions of AI agents will work alongside humans daily. He emphasizes that this shift marks a transition from an industry focused on tools to one centered around skills, with agents utilizing and collaborating with various tools to solve problems. Huang envisions a future where these agents can spawn new agents, collaborate, and create their own tools, significantly enhancing their capabilities as AI technology continues to evolve.
Huang highlights the importance of unsupervised learning in AI development, which allows models to learn from vast amounts of data without human labeling. He points out that the limitations of human involvement in data labeling have historically constrained AI’s growth. With advancements in unsupervised learning, AI can now generate and refine its own models, leading to exponential growth in capabilities. This shift is crucial as the available public data becomes increasingly limited, pushing companies to leverage proprietary data to differentiate their models.
The conversation also touches on the rapid advancements in computing power, surpassing Moore’s Law due to the rise of GPUs and parallel computing. Huang notes that AI’s ability to create software and improve itself is accelerating the development of new AI systems. As AI continues to evolve, it will increasingly automate software creation, leading to a future where AI writes code and potentially generates model weights, making traditional coding practices obsolete.
Huang emphasizes the need for effective onboarding processes for AI agents, drawing parallels to human employee onboarding. Just as new hires require training and context to perform effectively, AI agents will need similar support to understand their roles and tasks. This approach will enable AI to work more efficiently and intuitively, reducing the need for detailed prompts and allowing for smoother collaboration between humans and AI.
In conclusion, Huang expresses excitement about the transformative potential of AI and the new computing paradigm it heralds. He envisions a future where computing is dynamic and predictive, moving away from static software to real-time, context-aware solutions. This shift will require innovative hardware and software, and Huang believes we are on the brink of a significant evolution in how we interact with technology, making it an exhilarating time for advancements in AI and computing.