I remember reading a blog post saying that C. elegans is not a good example of neural networks, for the same reason demoscene is not a good example of computer programming. C. elegans connectome was tweaked and optimized by evolution to be as compact as possible.
So just like with demoscene, untangling and simulating it will require deep knowledge of the underlying "physical" platform. More complex organisms might actually be _easier_ to simulate.
A third of the body of C.elegans consists of neural tissue, it's incredibly energy-expensive. So it was optimized a lot, and there are some hidden interactions between neurons that are not directly connected.
Hormones, I think... I'm not sure exactly what the g.p. post is referring to, but in people there are a lot of non-direct interactions, too. The one's I'm (only slightly) familiar with are hormones. Maybe there are others?
"Hidden interactions" in c. elegans typically refers to chemical signaling rather than electrical. Of course, chemical signaling is tolerated in all nervous systems whenever its slow speed doesn't bring down the cruel hammer of natural selection.
The neuron state is impacted by: neurotransmitters, electromagnetic fields, the chemical soup that it sits in, etc.
An interesting example of how non-simple this stuff is: a new study (https://www.nature.com/articles/s42003-024-06778-2) where both the intracellular calcium level and the membrane potential were measured at the same time. The result:
The membrane potential encoded the presence of an odor
The intracellular calcium level encoded the magnitude of the odor
No worries, thanks for looking! I'm still new to learning about a lot of this stuff, and didn't realize neurons in the body could use chemical signals to communicate with the brain.
I still don't see how that indicates their brain is more advanced than we might think, though. What are the reasons to assume that?
my bystander understanding: the problem is that openworm knows how the neurons are connected, but not how they're weighted and trigger each other, or how it adjusts them, and it seems like there hasn't been enough funding to take it to the next level in the past, like, 15 years
it's unclear to me how much of the result from this paper is algorithmic rather than directly founded in measured data, but it's very cool to see more work in this area.
They claim their simulation moves in a "zigzag" pattern that's qualitatively similar to the real animal:
> During locomotion, the input of sensory neurons exhibited fluctuations due to the zigzag movement of the body’s head (Fig. 5c). The membrane potentials of each individual motor neuron also oscillated in response to the sensory input, especially the head motor neurons (Fig. 5d). The activation of muscle cells revealed traveling waves from the head to the tail (Supplementary Fig. 4), accompanied by alternating contractions and relaxations of dorsal and ventral muscle cells (Fig. 5e). These findings resemble observations in biological experiments
> They claim their simulation moves in a "zigzag" pattern that's qualitatively similar to the real animal:
It seems like they need to show it following chemical gradients the way c elegans does, I don't think any of these simulations has been successful at that.
loved that show! my favourite part is their dead eyes, knowing everything that is going to happen takes away all the joy and pain out of life, it's exactly like depression!
These animals are 'conscious' yet not capable of any kind of higher thought, basically just a early type of organic robot. Fascinating.