The Embodied Nature of Intelligence and the Flaws of the Intelligence Explosion Hypothesis

This podcast clip features François Chollet critically examining the widely discussed concept of an "intelligence explosion" in the context of Artificial General Intelligence (AGI). The intelligence explosion hypothesis posits that an AI, once capable of general problem-solving, could recursively improve its own algorithms, leading to an exponential and unchecked increase in its intelligence. Chollet challenges this notion by arguing that it relies on an implicit and flawed definition of intelligence, treating it as an isolated property of a brain, akin to the height of a building, rather than a complex emergent phenomenon.

Chollet proposes an alternative, more holistic understanding of intelligence, asserting that it emerges from the intricate interaction between a brain, a body (emphasizing embodied intelligence), and an environment. He contends that if any of these crucial components are missing or considered in isolation, a meaningful definition of intelligence cannot be formed. Consequently, merely tweaking an AI's internal algorithms or "brain" in isolation would not lead to an exponential increase in its capabilities, as the body and environment would inevitably become bottlenecks to the expression of that intelligence.

The discussion further delves into the concept of bottlenecks within complex, interdependent systems. Chollet explains that improving one part of such a system, like the brain in the brain-body-environment triad, will simply cause a new bottleneck to arise elsewhere. He suggests that even for highly intelligent humans today, their brain's raw processing power is often not the limiting factor for their expressed intelligence or achievements. Instead, the constraints often lie in the environment, the availability of significant problems, or the opportunities to apply their potential.

To illustrate this point, Chollet uses the analogy of "Einstein in his patent clerk days," suggesting that many individuals possess genius-level potential but may never express it without the fortuitous "meeting of great problem-solving capabilities with a great problem" at the right time and place. This highlights that intelligence is not merely a latent potential (like an IQ score) but its active manifestation through interaction with the world. This perspective implies that for AI to truly advance, it would require a co-evolution or co-improvement of its cognitive architecture, its physical embodiment, and the environment it operates within, rather than just isolated algorithmic self-enhancement.

"the problem of building such in the eye that itself is a problem that could be solved by your AI and maybe it could be so better than that and what humans can do so you're a I could start tweaking its own algorithm good that start being a better version of itself and so on iteratively in an in a recursive fashion and so you would end up with an AI with exponentially increasing intelligence"

"the notion of intelligence explosion uses an implicit definition of intelligence that doesn't sound quite right to me it considers intelligence as property of a grain that you can consider in isolation like the height of the building for instance right"

"intelligence emerges from the interaction between a brain a body like embodied intelligence and an environment and if you're missing one of these pieces then you can actually define interagency"

"it seems naive to think that our own brain is somehow the limit of the capabilities and this information is just I'm playing devil's advocate here this information processing system and then if you just scale it if you're able to build something that's on par with the brain you just the process that builds it just continues and it will improve exponentially"

"you cannot just tweak one part of this system back of this brain body environment system and expect capabilities like what emerges out of this system to just you know explode exponentially because anytime you improve one part of a system with many interdependencies like this there's a new bottleneck that arises"

"I don't think even today for very smart people their brain is not the bottleneck to the sort of problems they can solve"

"Einstein became Einstein because this was a meeting of a genius with a big problem at the right time right but maybe this meeting could have noon and ever happens and then I assume there's just been a patent clerk"

"intelligence is the meeting of great problem-solving capabilities with a great problem and if you don't have the problem you don't react spread intelligence all you're left with is potential intelligence"

"many people today are probably like genius level smart but you wouldn't know because they're not really expressing any of that was brilliant"