Real Intelligence Requires a Real Body

A situated agent is intelligent if its actions are appropriate for its environment, its goals, and its limitations. Intelligence is about appropriate action, not necessarily about reasoning or thinking. Most current AI paradigms are based on either 1) explicit representations of knowledge or 2) learning in artificial neural nets. Chatbots, such as Gemini and ChatGPT, are based on deep neural net Large Language Models. Those approaches are usually disembodied in that they do not physically interact with the world.

A different approach insists that embodiment is essential to intelligence. An embodied agent is simply one with a real body. Intelligence emerges, in evolution and individual development, through ongoing interaction and coupling of a physical body with a real environment. This paradigm underlies Embodied AI (EAI).

A robot is an artificial purposive embodied agent. EAI emphasizes the tight coupling of perception and action. Robotics is the ideal test domain for EAI. That motivated us to propose and build the world’s first robot soccer players as an Embodied AI challenge. The RoboCup soccer challenge has led to many experiments and theories for embodied multiagent real-time learning, decision-making and action.

Embodiment is an essential scientific requirement for intelligence. But it is also an engineering requirement in any application scenario that requires real-world interaction, such as a self-driving car, a factory robot, or the smart excavator and the semi-autonomous wheelchair that we developed.

Engineering Science, as a discipline, is an ideal preparation to work on EAI, given that EAI straddles Science and Engineering. I was lucky to get such a solid foundation in my undergraduate career at UofT in EngSci 6T6. The skills needed for work in EAI span the gamut including formal logic, epistemology, Bayesian probabilistic inference, algorithms, software and hardware architectures, sensor and actuator design, information theory and control theory. After I completed my BASc I explored many different paths to understanding and developing EAI. Great challenges await you!

Speaker bio

Alan Mackworth is a Professor Emeritus of Computer Science at UBC. He was educated at Toronto (B.A.Sc.), Harvard (A.M.) and Sussex (D.Phil.). He works on artificial intelligence with applications in constraint satisfaction, cognitive robotics, assistive technology, hybrid systems and constraint-based agents. He invented the world’s first soccer-playing robots. He has authored over 140 papers and co-authored two books: Computational Intelligence: A Logical Approach (1998) and Artificial Intelligence: Foundations of Computational Agents (2010; 2017, 2^nd Ed; 2023, 3^rd Ed.).

Alan co-founded the pioneering UBC Cognitive Systems Program, the UBC Centre for AI, Decision-making and Action (CAIDA), the AI network of BC (AInBC) and the IRIS and AGE-WELL NCEs.. He has served as President of AAAI, the world’s premier AI society.  He is a Fellow of AAAI, CAIAC, AGE-WELL, CIFAR and the Royal Society of Canada. He acts now as a consultant, writer and lecturer.