Evaluating AI Abilities
Our guest today is Konstantinos (Kozzy) Voudouris, a post-doctoral researcher at the Leverhulme Centre for the Future of Intelligence at the University of Cambridge. His research focuses on unifying artificial and natural intelligence, specifically in understanding how the common sense capabilities of animals are determined.
Kozzy discussed some of the latent capabilities in animals that are being studied. He focused on common sense capabilities such as object permanence. He used object permanence abilities as an example. He discussed the evolution cognition project and the major transitions that are key in the evolution of cognitive systems. He also addressed the overlap between AI and psychology.
Kozzy discussed animal AI. He explained how the trap tube test was used to understand animal’s causality in their world. He also discussed how he set up such experiments using reinforcement learning. He shared the deep learning and reinforcement learning libraries he used.
Kozzy also talked about the animal AI Olympics. Kozzy gave his thoughts on the trajectory of reinforcement learning and possibilities for more capabilities in the field. He also shared what was next for the project.
Learn more
Konstantinos (Kozzy) Voudouris
Konstantinos (Kozzy) Voudouris is a Research Associate at the Leverhulme Centre for the Future of Intelligence at the University of Cambridge. He works on the Major Transitions in the Evolution of Cognition project, focusing on identifying and characterising the stages in the evolution of neural systems that significantly increased the complexity of animal behaviour. He recently completed his PhD in Psychology, also at the University of Cambridge, focusing on the evaluation of cognitive capabilities in artificial intelligence. Since 2019, he has been heavily involved with the continued development of the Animal-AI Environment, a virtual laboratory for conducting psychological experiments on reinforcement learning agents. He has a background in cognitive science, linguistics, and the philosophy of science. You can find out more about the project and its history at www.animalai.org and you can use and contribute to it at https://github.com/Kinds-of-Intelligence-CFI/animal-ai
Thanks to our sponsors for their support
Georgia Tech Scheller College of Business
