张亚勤院士：基础大模型最终不超过10个，十年后机器人比人多

Core Insights - The new wave of artificial intelligence (AI) represents a fusion of information intelligence, physical intelligence, and biological intelligence, driven by unified token representation and scaling laws [3][6][24] - AI is transitioning from discriminative models to generative and reasoning models, with significant advancements in efficiency, cost-effectiveness, and open-source ecosystems [3][10] - Generative AI is rapidly evolving into intelligent agents, with task lengths and capabilities increasing, while associated risks are also amplifying [3][11] Industry Trends - The foundational large models are expected to converge globally to no more than ten, resembling operating systems in the AI era, leading to a new industrial structure of "base models + vertical/edge models + intelligent agent networks" [2][14] - The intelligent agent internet is anticipated to be the largest direction in the next 5-10 years and is considered a necessary path toward achieving Artificial General Intelligence (AGI) [5][24] - The development of intelligent agents is not merely a technological advancement but is also creating new economic forms and networks, fundamentally altering enterprise structures and product development [17][21] Future Projections - The transition from information intelligence to physical and biological intelligence is expected to be completed within 15-20 years, with significant milestones such as the emergence of Level 4 autonomous vehicles by 2030 [4][14] - The AI industry is projected to experience a scale of opportunity that is 100 times greater than previous technological eras, driven by the integration of foundational models, vertical models, edge models, and intelligent agents [22][24] - The intelligent agent internet is viewed as the next evolution of connectivity, moving from PC and mobile internet to a new paradigm focused on agent interactions [22][24]