Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

OpenWorm – A computational model of C. elegans worm

github.com

Post reply on HN
Print view View original

101–110 of 159 posts

Re: OpenWorm – A computational model of C. elegans worm

#101

Post by jacquesm »

Aardwolf wrote: post #97
burning_hamster wrote: post #60

Earlier quoted context omitted.

> "The neurons do not fire action potentials, and do not express any voltage-gated sodium channels." This an old and incorrect belief that largely derives from the difficulty of putting electrodes into their teeny, tiny neurons. Close relatives of C elegans that are larger (and hence more easily experimented on) do have action potentials, and for some neurons in C elegans, we also have good evidence of action potenti…

Given that we have a simulator of this worm right there (which includes it moving), can it really be up to debate whether it uses action potentials or not? I'd think the simulation has to get it right, and so needs to simulate action potentials if the worm has them, or not simulate them (but whatever the worm has instead) if not, right? Or could the simulation still be incorrect and only based on current assumptions,…

Not necessarily, because you could also simulate the worm without neurons at all. It's the closeness of the simulation to the real thing that demands that it is done right and the question effectively is: is this simulation close enough that if such a detail would be wrong that it would fail?

One way to answer that would be to add and remove such mechanisms to see if it would lead to different behavior.

Re: OpenWorm – A computational model of C. elegans worm

#102

Post by airstrike »

Maxion wrote: post #55
miohtama wrote: post #46

Earlier quoted context omitted.

Next: OpenSmoker.py

Lol imagine your IDE becoming addicted to smoking, and refusing to auto-complete or save files before you buy it more virtual cigarettes.

It's like when the xbox 360 would request users to drink Mountain Dew verification cans to validate

Re: OpenWorm – A computational model of C. elegans worm

#103

Post by _aavaa_ »

Aardwolf wrote: post #97
burning_hamster wrote: post #60

Earlier quoted context omitted.

> "The neurons do not fire action potentials, and do not express any voltage-gated sodium channels." This an old and incorrect belief that largely derives from the difficulty of putting electrodes into their teeny, tiny neurons. Close relatives of C elegans that are larger (and hence more easily experimented on) do have action potentials, and for some neurons in C elegans, we also have good evidence of action potenti…

Given that we have a simulator of this worm right there (which includes it moving), can it really be up to debate whether it uses action potentials or not? I'd think the simulation has to get it right, and so needs to simulate action potentials if the worm has them, or not simulate them (but whatever the worm has instead) if not, right? Or could the simulation still be incorrect and only based on current assumptions,…

That isn't immediately true, with enough fitting parameters you can capture the effect underlying behaviour without explicitly capturing it, or even without knowing it exists.

"With four parameters I can fit an elephant, and with five I can make him wiggle his trunk." - John von Neumann [0]

[0]: https://doi.org/10.1119%2F1.3254017

Re: OpenWorm – A computational model of C. elegans worm

#104

Post by chromanoid »

XorNot wrote: post #86
chromanoid wrote: post #73

Earlier quoted context omitted.

While the exact details were not understood, the mechanics were visible and explainable to some extent. The Wright brothers were at least inspired by birds. https://www.wright-brothers.org/History_Wing/Wright_Story/In... We still don't know how cells in C. elegans work together. It's neither visible nor explainable on a satisfying level.

No but we observably do know that that the connections between cells are important, to the point that by mimicking that we've derived significant benefit. The Wright flyer didn't flap, and the wings only superficially look like anything a bird has.

> No but we observably do know that that the connections between cells are important, to the point that by mimicking that we've derived significant benefit.

That is true and it shows even more how important observabilty is for science and engineering. That's also why a simulation that actually provides an accurate enough model of reality might help us so much. The problem with AI right now is, that we try or even claim to understand Unix by mimicking the functionality of transistors.

> The Wright flyer didn't flap, and the wings only superficially look like anything a bird has.

They tried to mimick bird wings when coming up with flight control mechanisms.

Re: OpenWorm – A computational model of C. elegans worm

#105

Post by mrguyorama »

ben_w wrote: post #98
bombolo wrote: post #94

Earlier quoted context omitted.

This makes no sense. Cells duplicate, but can you make a cell without splitting one in 2?

Why does it matter how it works? The 3.2 billion base pair model that is my genome doesn't understand the physics or the chemistry.

That's entirely wrong. The interactions of proteins and amino acids and ions IS, at a fundamental level, the evaluation of physics over time.

Re: OpenWorm – A computational model of C. elegans worm

#106

Post by api »

nightsd01 wrote: post #13

302 neurons doesn’t sound impressive to people who may be used to working with 7B+ parameter neural networks. But those neural networks have about as much in common with a biological neuron as a bicycle had with a horse. They can both travel pretty fast but one evolved naturally through over a billion years of harsh natural selection, and the other is a precisely tuned metal machine with a single purpose. Neurons are…

The best way of looking at it is that a single biological neuron is itself a complex machine full of genetic control circuits that sort of resemble neural networks and most importantly have memory/state that persists over both short and long periods of time. Each neuron is a full-ass living organism that itself is capable of learning and behavior, not a parameter is a model.

A virtual "neuron" by contrast is a very simple mathematical abstraction. It's vastly less computational complexity than a biological neuron. A connectome is only a very coarse grained map of how neurons relate, not a complete "neural network" layout. Not even close.

It might be possible to model a biological neuron using a sub-neural-network with state within a larger neural network, but assuming that can be computationally equivalent we don't really know how many equivalent computational "neurons" would be required to model the full breadth of computationally relevant biological neuron behavior.

So a worm with 302 biological neurons could be computationally equivalent to billions of virtual neurons. We really don't know.

Given that neurons have memory it may look a little like LSTM networks, and biological neural networks are not just feed forward so they're definitely closer to an RNN.

The above is why I laugh at the mind uploading people and would only stop laughing if we could both understand and model the relevant behavior of biological neurons and somehow extract usable state from living neurons. That's all 100% science fiction at the moment. The people who think we are about to upload minds are ignorant of biology.

Re: OpenWorm – A computational model of C. elegans worm

#107

Post by PTOB »

Gordonjcp wrote: post #49
dilawar wrote: post #33

Earlier quoted context omitted.

I did my Ph.D. in computational neuroscience (2020 grad). This project is often used to joke about the limitation of computational modelling of nervous system. If you can't compute the behavious of an effing worm with mere ~300 neurons, whats the point of all hot-air around connectomics (mapping connections of the brain). Connectomics used to be a big word when I started my Ph.D.. The apologists are always like, "rea…

> Connectomics used to be a big word when I started my Ph.D. On a scale of one to ten, how sick are you of people asking you about Neal Stephenson - "The Fall, or Dodge in Hell"?

Plot twist; he's got an "Ask Me about Neal Stephenson" t-shirt.

Re: OpenWorm – A computational model of C. elegans worm

#109

Post by bee_rider »

petters wrote: post #22
nightsd01 wrote: post #13

302 neurons doesn’t sound impressive to people who may be used to working with 7B+ parameter neural networks. But those neural networks have about as much in common with a biological neuron as a bicycle had with a horse. They can both travel pretty fast but one evolved naturally through over a billion years of harsh natural selection, and the other is a precisely tuned metal machine with a single purpose. Neurons are…

> billions of DNA base pairs controlling their behavior Total genome size of C elegans is 100M.

100Mdna, and they call it elegant? Biologists must be VSCode users.

Re: OpenWorm – A computational model of C. elegans worm

#110

Post by bee_rider »

anotheryou wrote: post #108

And to me this is just as sentient (or not) as the physical worm and deserves worm rights!

Does this actually become an interesting question at this point? I guess worms don’t have many rights anyway, but if they did, and this was a totally accurate simulation, why not give the digital worm rights?
Post reply on HN