Abstract: The next generation of artificial intelligence will not be engineered as static artifacts, but will emerge as autonomous ecosystems capable of self-directed growth, continual adaptation, and open-ended evolution. In this talk, a framework for Self-Evolving Autonomous AI Systems is presented, in which intelligence arises from the symbiotic interaction between heterogeneous agents and the algorithms that govern them, driven by large-scale evolutionary computation.

At the core of this framework lies a simple yet powerful principle: intelligence should not be embodied in a single monolithic model, but distributed across a population of cooperating multi-agent systems. Supported by large-scale, scalable evolutionary infrastructure, these agents interact through dynamic graph structures to transfer knowledge, specialize, reorganize, and merge in response to new tasks. Beyond model evolution, the algorithms responsible for learning, coordination, and adaptation are themselves subject to meta-level evolution, enabling the system to continually redesign how it learns and reasons.

This talk unfolds in three stages. First, the conceptual and mathematical foundations of self-evolving intelligence will be presented, framing autonomous AI systems as co-evolving populations of models and algorithms governed by multi-level evolutionary dynamics. Second, I will demonstrate how recent advances provide essential building blocks for this paradigm, including accelerated evolutionary computation, evolutionary model merging, and automated algorithm designs. Finally, some open challenges for realizing autonomous AI ecosystems will be discussed, such as stable long-term evolution, verifiable algorithmic safety, and adaptive protocols that respect hardware and privacy constraints.

Biography: Professor Kay Chen Tan is the founding Head and Chair Professor of Computational Intelligence at the Department of Data Science and Artificial Intelligence at The Hong Kong Polytechnic University. He has co-authored eight books and published over 300 peer-reviewed journal articles in the fields of computational intelligence and evolutionary computation. His works have collectively been cited more than 40,000 times, and he has an h-index of 100. His contributions to the scientific community have garnered him various accolades, including being named an IEEE Fellow and a Hong Kong RGC Senior Research Fellow. Additionally, he has consistently ranked among the World’s Top 2% Most-Cited Scientists by Stanford University and has been recognized as a Highly Cited Researcher in 2024 and 2025 by Clarivate.

Professor Tan has held numerous key editorial and leadership positions, including Chair of the IEEE CIS Fellow Evaluating Committee (2027) and Vice President for Publications (2021–2024) of the IEEE Computational Intelligence Society (CIS). He served as Editor-in-Chief of the IEEE Transactions on Evolutionary Computation (2015–2020) and the IEEE Computational Intelligence Magazine (2010–2013), and currently acts as Chief Co-Editor of the Springer Book Series on Machine Learning: Foundations, Methodologies, and Applications.

Professor Tan has been recognized with various awards, such as the IEEE CIS Evolutionary Computation Pioneer Award (2026), the IEEE CEC Best Paper Award (2025), the IEEE CAI Best Paper Award (2024), the IEEE Computational Intelligence Magazine Outstanding Paper Award (2024, 2019), the IEEE Andrew P. Sage Best Transactions Paper Award (2020), the IEEE Transactions on Neural Networks and Learning Systems Outstanding Paper Award (2016), and the IEEE CIS Outstanding Early Career Award (2012). Beyond research excellence, Professor Tan has delivered over 80 plenary and keynote lectures and co-organized more than 60 international conferences, including his roles as General Co-Chair for the 2019 IEEE Congress on Evolutionary Computation and the 2016 IEEE World Congress on Computational Intelligence.