Agentic AI systems represent a crucial step towards realizing AGI, where AI can perform any intellectual task that a human being can. These systems are designed to operate with a degree of independence, making decisions and taking actions without human intervention, based on their programming, learning, and the data they can access. The concept of agency in AI revolves around the capacity of these systems to understand their environment, plan actions towards achieving specific objectives, and adapt to changes or unforeseen circumstances. This involves complex decision-making algorithms, learning mechanisms, and the ability to interact with the environment in a meaningful way. The development of agentic AI systems pushes the boundary of AI from narrow, task-specific applications to more generalized and adaptable systems that can operate in a variety of contexts and perform multiple tasks.
The idea of machines or software systems acting with a level of agency has been a part of science fiction and theoretical discussions since the early days of computing. However, practical research and development in this direction have significantly accelerated in the 21st century, especially with advancements in machine learning, robotics, and natural language processing technologies.
While it's challenging to attribute the development of agentic AI systems to specific individuals due to its interdisciplinary nature, pioneering work by researchers in the fields of machine learning, robotics, and cognitive sciences has been crucial. Figures such as Geoffrey Hinton, Demis Hassabis, and Yann LeCun have made significant contributions to the underlying technologies that enable the development of agentic AI systems.