Back to How Do We Learn Dec 2025 Aliveline

Day 8

Alright first day I feel normal, no melatonin / weird sleep!

If we accept the universality picture – “macroscopic universality, microscopic degeneracy” – the implication is that DNA is encoding something that across brains can do various different things at the microscopic level but end up doing the same thing at the macroscopic/statistical level. That is the realm of complex systems. Or more specifically, it’s like specifying local rules that are constraining enough that different ‘coordinate systems’ can emerge at the local level while still having the same macro behavior. How might you actually do that? What systems do we know of today that exhibit such properties?

I guess in some sense even the concept of an ideal gas being made up of tons of microscopic unobservable degenerate gas particles which end up in a particular macro state is similar? Pretty vague analogy.

Now that there are large-scale brain recordings, I did sense some tension between the top-down populations of neurons view vs the bottom-up single neurons approach in computational neuroscience. It seems like the top-down view is much more powerful and popular nowadays, of course only enabled due to advances in recording technology. If you look at it from a DNA lens, it’s likely that DNA can’t be encoding what every single neuron does – that’s a very inefficient way to achieve the macro level behavior that matters, and doesn’t match recent work. Also natural selection acts on the macro behavior level, so that must be the level DNA works at. But it does somehow have to bridge the micro to the macro level. How does it do that??

I wonder if the evolutionary history of C. Elegans can shed some light on this? Or if you grow a single, say cortical, neuron in culture what does it do?

Convo with ChatGPT on the above ended with this idea:

Can we treat developmental / genetic parameters as the small-scale couplings, and macroscopic neural manifolds as fixed points or order parameters of an RG-like flow?

Convo with ChatGPT where I was wondering if I could use C. Elegans data to show a phase transition in the populations of neurons that fire when the organism understands a rule, and which datasets could help show this.