Y Combinator Startup Podcast #14 - Ex Machina's Scientific Advisor - Murray Shanahan
Jun 28, 2017
Murray Shanahan, a Research Scientist at DeepMind and a professor at Imperial College London, shares intriguing insights as a scientific advisor for Ex Machina. He discusses the evolution of AI from classical approaches to modern machine learning, emphasizing the importance of embodiment in intelligence. Shanahan reveals his experiences advising on the film's realism and highlights the influence of philosophical texts on character development. Additionally, he explores the stark contrasts between Hollywood portrayals of AI and the realities of current advancements.
AI Snips
Chapters
Books
Transcript
Episode notes
Frame Problem Shapes AI Trajectory
- Murray Shanahan traced his PhD work to optimizing Prolog-style logic queries by storing relationships to avoid recomputation.
- He links that early work to the recurring 'frame problem' of focusing on relevance in cognition and AI.
Relevance Is Central Across AI Approaches
- Shanahan says the frame problem recurs across symbolic AI, neuroscience, and modern machine learning as focusing on relevance.
- Contemporary ML must also learn to ignore irrelevant features to solve tasks efficiently.
Embodiment Grounds Intelligence
- Shanahan argues human intelligence originates in embodiment: brains evolved to navigate and manipulate 3D space.
- Cognition and consciousness are rooted in bodily interaction with the world, not just disembodied computation.
