Tomaso Poggio on Brains, Minds, Machines, and the Nature of Intelligence

Tomaso Poggio, a distinguished professor at MIT and director of the Center for Brains, Minds, and Machines, discusses the profound problem of intelligence, deeming it the greatest challenge in science. He recounts his early inspiration from Albert Einstein's thought experiments and the power of the mind to uncover physical reality, which led him to pursue artificial intelligence as a means to amplify human problem-solving capabilities. Poggio expresses optimism about machines eventually surpassing human intelligence, while acknowledging the critical need to precisely define terms like "intelligence" and "consciousness." He emphasizes the historical and ongoing role of neuroscience in driving AI breakthroughs, citing the biological origins of concepts like reinforcement learning and deep learning.

A key distinction Poggio highlights is the contrast between current deep learning's reliance on "big data" (millions of labeled examples) and biological learning's efficiency, often operating on an "n going to one" principle, where learning occurs from very few examples. He delves into the nature vs. nurture debate concerning brain development, particularly the rapid specialization of regions like the face recognition area. Poggio hypothesizes that evolution might encode "weak priors" or highly plastic brain areas designed for quick imprinting based on early experience, rather than hardwiring detailed neural circuits. He also contrasts the modular, engineered separation of hardware and software in traditional computers with the more intertwined and complex levels of abstraction in the brain, arguing that a comprehensive understanding of the brain necessitates interdisciplinary collaboration across multiple levels.

The conversation offers practical insights for AI development, suggesting that future progress will likely stem from continued inspiration from neuroscience, particularly in developing more data-efficient learning algorithms. Poggio introduces the concept of "compositionality" as a fundamental structural property of functions that deep networks excel at learning, explaining their success in domains like vision and language. This understanding guides the design of effective neural network architectures. His work underscores the necessity of interdisciplinary approaches, integrating insights from physics, neuroscience, and computer science to tackle the intricate problem of intelligence.

Broader implications of the discussion touch upon the philosophical underpinnings of AI, questioning the ultimate limits of machine capabilities and the very definition of intelligence. Poggio's perspective on the brain's compositional structure, potentially shaped by evolutionary constraints such as the difficulty of growing long-range connectivity, provides a compelling explanation for why certain problems are inherently "easy" for biological and artificial systems, while others remain elusive. The ongoing quest for alternative optimization algorithms beyond stochastic gradient descent further highlights fundamental open problems in both AI and neuroscience, positioning the understanding of intelligence as a foundational scientific endeavor with far-reaching consequences for humanity's future capabilities.

Einstein was a hero to me and I'm sure to many people because he was able to make of course a major major contribution to physics with simplifying a bit just a Gedanken experiment a fourth experiment...

So there is a lot to be said about you know trying to be to do the opposite or something quite different from what other people are doing that's actually true for the stock market never never buy for very bodies by and also true for science.

I don't see in principle why computers at some point could not become more intelligent than we are although the word intelligence it's a tricky one and one who should discuss which I mean with that in intelligence consciousness yeah words like love is all these are very you know you need to be disentangled.

What if we could find the key to an intelligence you know ten times better or faster than Einstein so that's sort of seeing artificial intelligence as a tool to expand our capabilities...

My personal bet is that there is a good chance they continue to play a big role maybe not in all the future breakthroughs but in some of them at least in inspiration.

One of the main differences and you know problems in terms of deep learning today and it's not only deep learning and the brain is the need for deep learning techniques to have a lot of labeled examples.

I think the biological world is more n going to one Hey a child can learn the beautiful wrote a very small number of you know labeled examples like you tell a child this is a car you don't need to say like imagenet you know this is a car this is a car this is not a car this is not a cat 1 million times.

For the cortex I think your question about hardware and software and learning and so on it's it I think is rather open and you know it I find very interesting for easy to think about an architecture computer architecture that is good for vision and the symptom is good for language seems to be you know so different problem areas that you have to solve but the underlying mechanism might be the same.

If the function is made up of functions of function so that you need to look on when you are interpreting an image classifying an image you don't need to look at all pixels at once but you can compute something from small groups of pixels and then you can compute something on the output of this local computation and so on that is similar to what you do when you read the sentence you don't need to read the first and the last letter but you can read syllables combine them in words combine the words in sentences so this is this kind of structure.