Machine Learning and Speech Recognition with Kai-Fu Lee

Tom sits down with Bruce Buchanan, a PhD Philosopher turned machine learning researcher.  Bruce produced a key milestone for machine learning in the 1970s by creating the first program that discovered new symbolic knowledge publishable in a scientific journal.Bruce has held professorships at the University of Pittsburgh (Philosophy and Medicine) and Stanford University (Computer Science).Tom Mitchell is the Founders University Professor at Carnegie Mellon University. Produced by the Stanford Digital Economy Lab.

Tom chats with John Laird, who has spent the past 40 years trying to build an AI agent that accomplishes the full range of human cognitive abilities, beginning with his 1980s PhD research on the SOAR model of human cognition with Allen Newell and Paul Rosenbloom.John E. Laird received his Ph.D. from Carnegie Mellon University in 1985, and is John L. Tishman Emeritus Professor of Engineering at the University of Michigan. He is one of the original developers of the SOAR architecture and leads its continued development and evolution. He was a founder of Soar Technology. He is a AAAI, ACM, AAAS, and Cognitive Science Society Fellow. In 2018, he was co-winner of the Herbert A. Simon Prize for Advances in Cognitive Systems.

Tom meets with Kai-Fu Lee, a pioneer in using machine learning to significantly advance speech recognition.Kai-Fu, former president of Google China and now Chairman of Sinovation Ventures and CEO of 01.AI, has led speech, machine learning and AI efforts at several top firms, and is now one of the top AI venture capitalists in China.

What is the relationship between neural network approaches in machine learning, and real neural networks in the brain? Today's guest Jay McClelland is a cognitive scientist who has spent decades studying this question. Jay is Lucie Stern Professor of Psychology and (by Courtesy) of Linguistics and Computer Science and Director of the Center for Mind, Brain, Computation and Technology at Stanford University. He discusses his 50 year journey modeling cognition in the brain with artificial neural networks, and his role in the 1980s emergence of neural networks in machine learning.

Tom interviews Daphne Koller, a Stanford professor turned serial entrepreneur. Daphne is widely known for her research at the intersection of machine learning and probabilistic reasoning.Daphne is a member of the U.S. National Academy of Engineering, and is currently CEO of Insitro, a company at the intersection of machine learning and human biology.

Tom meets with Dr. Dean Pomerleau, who as a CMU PhD student in the 1980s was the first person to demonstrate that a neural network could be trained to automatically steer a self-driving vehicle.Dean's results shocked the research community, and paved the way for decades of follow-on research leading to today's self-driving cars.

Tom sits down with Michael I. Jordan, Director of Rearch at Inria and Professor Emeritus of the Departments of EECS and Statistics, University of California, Berkeley. Michael has been a major contributor to machine learning, especially at the intersection of statistics and machine learning.Michael discusses his research trajectory, including how it has been inspired by ideas from control theory, statistics, and most recently economics.

What would a "theory" of machine learning tell us? In this episode Tom meets with the person who invented what is now the widely accepted definition of supervised machine learning: Turing Award recipient and Harvard Professor Leslie Valiant.Leslie tells us how he got interested in the problem, his contribution, the evolution of machine learning theory over the decades, and his advice to new researchers.

Tom speaks with Ross Quinlan, whose algorithms C4.5 and ID3 helped establish decision trees as one of the most popular approaches in machine learning, and who founded RuleQuest Research, which accelerated the commercial adoption of machine learning.Ross (published as "JR Quinlan") describes a sabbatical visit to Stanford University where he took a course that drove him to invent the first successful learning algorithm for decision trees, follow-on research that led to decision trees becoming one of the most popular machine learning algorithms, and his experience moving from academia into the commercial world.

Tom interviews Rich Sutton, Research Scientist at Keen Technologies, Professor of Computing Science at the University of Alberta and co-winner of the 2024 ACM Turing Award for his foundational research on reinforcement learning.Rich discusses why the common framing of machine learning as 'supervised learning' is insufficient, and how reinforcement learning reframes the problem. He discusses how reinforcement learning has developed as a subfield of machine learning, the influence of Harry Kopf on his early thinking, his long-time collaboration with Andy Barto, his views about today's state of the art, and more.