Research and Prospects on the Development of Agents in the Era of Large AI Models
This paper reviews the development history of large AI models,tracing the evolution from early expert systems to current models like GPT-4 and Claude-3,which are based on the Transformer decoder architecture.It discusses the breakthrough advancements in natural language processing brought about by these models.The paper also outlines the development trajectory of agents,from simple reflexive agents to goal-based agents and multi-agent systems.Furthermore,it emphasizes how large AI models empower the development of agents by enhancing task understanding,improving decision-making and reasoning abilities,and boosting human-computer interaction efficiency.Finally,it explores the future trends in agent development in the era of large AI models,including embodied intelligence,autonomous learning,personalized intelligence,and the challenges and opportunities related to ethics and safety.
AI macromodelingmacrolanguage modelingintelligentsiahuman-computer interactionnatural language processing
万方数据
