On January 21 (local time), Jensen Huang, the CEO of NVIDIA, highlighted at the Davos Forum that AI models have witnessed three significant breakthroughs over the past year. Firstly, AI models have transitioned from being prone to generating false or misleading information (hallucinations) to being applicable in research domains. This is exemplified by the emergence of agentic AI, which can reason, plan, and answer questions without requiring prior training in the relevant fields. Such AI demonstrates a remarkable leap in autonomous cognitive capabilities.

Secondly, the introduction of DeepSeek, the inaugural open - source reasoning model, has spurred the development of a vibrant ecosystem of open - source reasoning models. This has empowered a multitude of companies, research institutions, and educators to harness open - source models for their respective endeavors, promoting widespread innovation and collaboration.

Thirdly, substantial advancements have been made in physical AI. This type of AI can now comprehend not only language but also the physical world. Its scope of understanding encompasses various fields such as biological proteins, chemistry, and physics, as well as more specialized areas like fluid dynamics, particle physics, and quantum physics, marking a profound expansion of AI's cognitive reach.