Huang announced three revolutionary shifts in the development of artificial intelligence

At the Davos Forum, Nvidia CEO Jensen Huang presented the most important milestones in the development of AI over the past year, which, according to him, have radically changed the perception of the capabilities of modern models. It’s not just about improving performance, it’s about a qualitative transition to fundamentally new applications and capabilities of artificial intelligence.

From demos to functional solutions

Jensen Huang noted that until recently, AI models were prone to hallucinations and could not be reliably applied in real tasks. However, over the past year, there has been a significant breakthrough — models have learned not only to reproduce information, but also to demonstrate genuine cognitive abilities. They began to show reasoning skills, long-term planning, and the ability to answer complex questions without prior training on specific scenarios.

It is these capabilities that formed the basis of a new class of solutions – the so-called Agent-based AI, which can act independently, make decisions and adapt to new situations. This means moving from simple chatbots to full-featured, intelligent assistants capable of performing complex, multi-step tasks.

Democratizing Models: The Open Source Era

The second significant achievement, according to the head of Nvidia, is related to the flourishing of the ecosystem of open source models. The tipping point came when the open-source functional inference model, DeepSeek, first appeared, which showed that high-performance AI solutions were no longer the monopoly of large corporations.

This breakthrough has created an ecosystem of accessible models that can be implemented and improved by companies of all sizes, research labs, and educational institutions. This has led to accelerated innovation and allowed researchers and developers around the world to experiment with their own AI applications based on proven models. Openness and accessibility are no longer an obstacle to the introduction of artificial intelligence.

Physical AI: Understanding Reality, Not Just Words

The third direction of progress, identified by Jensen Huang, represents a fundamentally new stage in the development of AI — the transition from purely linguistic models to systems capable of understanding and interacting with the physical world. Physical AI is not limited to text processing; He is aware of the laws of nature and can work with biological processes, chemical reactions, and physical phenomena.

In particular, the models have demonstrated the ability to analyze the dynamics of fluid flow, calculate the behavior of elementary particles, and even work with the concepts of quantum physics. This opens up unprecedented opportunities for scientific research, the development of new materials and the solution of complex engineering problems that previously required the involvement of human experts.

Thus, says Huang, AI is evolving from an information-processing tool to a full-fledged human partner in cognition and transformation of the surrounding reality.