Nvidia CEO Jensen Huang highlighted three major advancements in AI models over the past year during his speech at the Davos Forum. According to Odaily, Huang noted that initially, AI models exhibited numerous illusions, but last year marked a significant shift as these models began to be applied in research fields. They demonstrated capabilities such as reasoning, planning, and answering questions without specific training, leading to the emergence of Agentic AI.
Huang identified the second breakthrough as the introduction of open-source models, with the launch of the first open-source inference model, DeepSeek, being a pivotal event for various industries and companies. Since then, the ecosystem of open-source inference models has flourished, enabling companies, research institutions, and educators to leverage these models for various applications.
The third area of significant progress is in physical AI, which not only comprehends language but also understands the physical world, including biological proteins, chemistry, and physics. In the realm of physics, AI has shown the ability to grasp concepts such as fluid dynamics, particle physics, and quantum physics.