I recently came across a podcast series on AI and robotics called Talking Robots, put together by the Laboratory of Intelligent Systems, EPFL, Switzerland. They interview people in the field on a wide variety of topics related to robotics. I just finished listening to the interview with Barbara Webb, who published an excellent paper in 2001 called Can robots make good models of biological behaviour?

She definitely seems to be in the camp that favors embodied cognition, a school of thought that thinks the best way to understand cognitive systems is to build ones that interact with the real world. She's in the process of trying to build a robot that replicates aspects of behavior of crickets.

One thing that she said in the interview that struck me was that she was interested in general in how things think, and she thought it would be a good strategy to focus on simple neural systems like those of insects to learn how they do what they do before trying to tackle the immense complexity of the human (or even other primate) brains.

I used to think this way. I think there's a lot to be learned at the neural level that maps from smaller, simpler brains to larger, more complex ones. But at levels above the neuron, I think the organization and function of the neural systems of insects and primates massively diverges.

For starters, while insect behavior can be extremely flexible at times, most of that behavior is still hard-wired. Insects are not fundamentally learning creatures. Their behavior can be adaptive, cooperative, and extremely complex, especially in groups, but they do not learn very much through the course of their lives.

I believe mammals brains, the neocortex in particular, evolved as a general-purpose learning substrate. It is highly uniform, regular, and plastic. We are readily able to learn a vast array of temporal patterns, including skills and knowledge, that other species simply are not. And I think this is because our brains are not just quantitatively, but qualitatively different.

So I'm doubtful that studying cricket brains is going to scale up to human brains, even incrementally. Still, I admire her work, and if you have an interest in such things, you should check out the rest of the podcast series.