Prediction at the Core of both Natural and Artificial Intelligence

Downing's talk is organized as part of the 2025 Norwegian AI Society (NAIS) Symposium. For more information about the symposium, go to NAIS' web pages.

Abstract

Prediction is a cognitive advantage like few others, inherently linked to our ability to survive and thrive. Our brains are awash in signals that embody prediction, across a wide range of temporal and spatial scales. In this lecture, I will investigate the origins and anatomy of natural and artificial neural networks as predictive mechanisms. In addition, links to nascent large-language models (LLMs) will be hard to avoid, since their underlying intelligence appears to emerge from performing simple predictive tasks, albeit with complex attention-based representations and massive transformer architectures.

Biography

Keith Downing is a professor of Artificial Intelligence (AI) at the Norwegian University of Science and Technology (NTNU) in Trondheim, Norway. He is also an Affiliate Scholar in the Neuroscience Department of Oberlin College. His main interests, experiences and publications are at the intersection of AI, Artificial Life (ALife), and Computational Neuroscience. He spends most of his time either a) teaching several of NTNU’s AI classes, b) programming neural networks and evolutionary algorithms, or c) lecturing to the general public on technical and ethical aspects of AI. Keith's AI and ALife research culminated in two books: "Intelligence Emerging" (MIT Press, 2015) and "Gradient Expectations" (MIT Press, 2023).