17. Bruno Olshausen, PhD on how the brain compresses, sees, and makes sense of the world

In this episode of Sentience, Bruno Olshausen, PhD, professor at UC Berkeley, tells the story of how he went from wanting to build brain-like robots to uncovering one of neuroscience's most surprising results. Early in his career, he and a collaborator showed that if you train a simple computer model to represent images as efficiently as possible, it spontaneously reinvents the same kind of visual "building blocks" found in real brains—a discovery almost nobody believed at first. Olshausen walks us through why the brain seems to prefer having most of its neurons sit quietly at any given moment, why this looks so different from how modern AI vision systems work, and why brains might not be "efficient" so much as bound by the same physical limits as everything else. We also get into why the brain sends information as sudden spikes rather than smooth signals, and what all of this might mean for building truly autonomous machines.Timestamps(00:00) – Welcome to Sentience and meeting Bruno Olshausen(00:41) – From building robots to studying the brain(03:37) – An early chapter at NASA(05:56) – Why recognizing objects is harder than it sounds(09:45) – Where AI vision and real vision part ways(17:48) – A brief history of "efficient" brains(32:46) – The experiment that surprised everyone(36:04) – Why nobody believed it at first(46:17) – Making sense of uncertainty: how the brain guesses(54:47) – Why most of your neurons are doing nothing right now(01:07:16) – Two very different ways the brain stores memories(01:15:01) – What "efficient" really means for a brain(01:22:33) – Why the brain uses spikes instead of steady signals(01:28:19) – What Bruno's chasing nextBooks referenced:Sterling, P. & Laughlin, S. Principles of Neural Design. MIT Press.MacKay, D.J.C. Information Theory, Inference, and Learning Algorithms. Cambridge University Press.Core papers referenced:Olshausen, B.A. & Field, D.J. (1996). "Emergence of simple-cell receptive field properties by learning a sparse code for natural images." Nature, 381, 607–609.Barlow, H.B. (1961). "Possible principles underlying the transformation of sensory messages." In Sensory Communication.Rozell, C.J., Johnson, D.H., Baraniuk, R.G., & Olshausen, B.A. (2008). "Sparse coding via thresholding and local competition in neural circuits." Neural Computation, 20(10), 2526–2563.