Over at LessWrong niconiconi recently posted a somewhat pessimistic assessment of the last 10 years of progress on emulating C. Elegans, including OpenWorm:
Recently I looked into this subject and talked with a postdoc researcher. It is interesting to see this discussion on HN.
C elegans is far more complicated than the connectome abstraction used in these projects. There are all kinds of ion channels, including new ones recently discovered (https://pubmed.ncbi.nlm.nih.gov/34388373/ ), that are regulated by genes and other physical/biochemical factors in the body that are very poorly understood. We only have slightly better understanding of its brain development recently ( https://www.nature.com/articles/s41586-021-03778-8 ). The in vivo research on C. elegans neural system mostly uses the concentration of Ca+ as the neuron activity indicator, which lefts out other details.
According to the postdoc that I talked with, we are still very very far from even accurately understanding a single neuron, let alone simulating any neural network completely.
C. Elegans is the perfect first candidate for understanding the whole brain of some animal.
It doesn't even have spiking neurons. Whole Brain Emulation of merely 302 neurons would not be that hard if we understood how worm brain and neurons work.
Replicating limited outward behaviour with simplified neuron model.
vs.
Fully simulating the functioning of the neuron so that it does not replicate some observed behaviors, but simulates the brain function accurately at the neuron level.
Hodgkin-Huxley neurons are fully functioning neurons as far my understanding goes, furthermore, if not mistaken openwarm simulator has ability to simulate hydrodynamics on cellular levels.
I always thought it would be much less interesting to simulate a worm on a cellular level, and much more interesting to simulate some single-cell organism on an atomic or even subatomic level. The latter feels more like “real life”, IMO.
I don’t know which one is more difficult, though.
Lots of interesting ethical questions. In the future we might be able to simulate a human on that level. Is it really alive? Does it have rights? If we can “scan” someone and simulate them virtually, are they the same person? If the physical body dies but there’s still a digital copy, can we say they’re still alive?
> In the future we might be able to simulate a human on that level. Is it really alive? Does it have rights?
Purely probabilistically, we are such humans. If any 'base' universe can easily simulate humans, for fun or research, such simulations will soon vastly outnumber the 'real' people. And because of that, we have to assume we are in an ancestor simulation.
I really disagree with the fundamental assumption there.
I find it very difficult to believe that a base universe could ever simulate something of the complexity of our universe without massive and pointless expense, at best. And even if they could, I find it even more unlikely that simulated universes would ever even be dense in the subset of multiverses (whichever flavor your like) which contain something like humans.
I agree, and additionally, it might take an awful lot of time until a civilization is advanced enough to reach the point where it can simulate other beings. For all we know, we can be the first beings to ever reach this point.
> If we can “scan” someone and simulate them virtually, are they the same person? If the physical body dies but there’s still a digital copy, can we say they’re still alive?
At the absolute best, you have a poor copy of the person.
We have had examples of poor copies forever. We call them twins. If you have a twin, that twin isn't you, and you're not them.
The most interchangeable your copy could be with you would be that the copy might be able to log into and post from your online accounts and fool many of your followers, if you have any; or to withdraw your money from your bank account and buy the sort of things you might have bought if you still had the money.
It would still be a copy, and not you. The copy could fool other people into thinking it was you, as twins have done to one another. Its motivations would, necessarily, instantly diverge from yours, e.g. preferring to spend your money on feeding and housing itself, over feeding you.
Well, they would be (in theory) indistinguishable from the physical person. You’d be able to interact with them, and their thoughts, feelings, memories etc would all still be there. In many ways the “soul” is still there. Obviously this is very much a matter of opinion.
It would be a copy. You might choose to give it the original's property, on the assumption that the original would have wanted that, or on the basis of a written will that actually says so. Then, the copy could carry on as if it were the original.
And, in fact, we do that continuously. Every day you wake up believing yourself to be the person who went to bed the night before. It is close enough.
> While our ultimate goal is to simulate every [worm] cell in C. elegans, we are starting out by building a model of its body, its nervous system, and its environment. (...) To get a quick idea of what this looks like, check out the latest movie: https://www.youtube.com/watch?v=SaovWiZJUWY
I mean simulating muscle contractions treats it like a clockwork machine; that's not very impressive. If you go deeper you can look at internals of cells, metabolism, chemical energy, protein interactions. That's how the worm works.
This is like simulating a computer by using a model of RAM and CPU, when you really need to model electricity & magnetism (and stray gamma rays) for a true simulation.
Ah I see... so it sounds like I misunderstood. You're calling for a higher-fidelity physical model, not something metaphysical :)
I still think it's impressive if we can get a decent-fidelity model of C. elegans behavior at the project's current level of abstraction.
I wonder how much behavioral difference we'd see from a lower level of simulation? After all, you can do a lot with a "RAM and CPU" level simulation of a computer...
I think the example is not great. The problem is not that we could be doing better, but rather that without a more granular view, the model will simply not work.
More like trying to model the pixels on a screen without accessing the underlying program that's making decisions on what to show.
On that topic: it can't necessarily be ruled out that they might have a soul. (In the sense of possessing consciousness.) Less likely the fewer degrees of freedom you have, and certainly seems like a long shot for ~300 neurons, but we don't know where the lines may be drawn and we may not know for a very long time.
I'm noncommittal toward the prospect that a C. elegans might be conscious (though I doubt it). But here I was using "soul" to mean something metaphysical, requiring something beyond the physical configuration of those ~300 neurons to explain.
It's impossible to really rule out such a thing but at the same time I don't think there's any way we can understand how to actually simulate or even identify that.
A soul is not simply consciousness. A soul usually means some non-physical entity that can exist independently from the physical body (e.g. when you die).
How do you know it is pure nonsense? We actually have no scientific evidence either way. There is a reason why discussing the "hard problem of consciousness" is so popular. There are many people now advocating for a universal consciousness, i.e. that all space time has intrinsically a mind-like element that is merely "tapped into" by physical bodies.
It's foolish to pretend that it's a solved problem when it's not. At the present moment we don't actually have a solid idea either way.
>We actually have no scientific evidence either way.
That's how we know it's nonsense. When there's no evidence for something it doesn't mean that it's 50% Likely to be true.
If I told you there's a dragon in my garage you should immediately think that I'm lying because you've never seen any prior evidence of dragons anywhere.
Mathematically, things with no evidence for their existence don't exist.
Sagan’s Dragon doesn’t really apply here, though. Because, while we don’t have evidence for something non-physical behind mind, we do have a phenomenon - conscious subjective experience - occurring with no real physical explanation anywhere in sight. Or even conceivable, really. Chalking it up to a specific religious idea of a soul or anything is a bridge too far, of course, as is believing anything one thing with more fervor than just a hunch, but...
Well, the analogy falls apart because whether there’s a dragon in the garage really is something physically verifiable, and it’s hard to know if where the there that’s there comes from is. But it’s a little like if you occasionally heard loud stomping noises or roars coming from your garage, and things in there were turning up with big bites taken out of them or singed with fire. No matter how much you look you can’t find the cause. Is that a dragon? Dragons don’t exist...but who knows?
Yeah I definitely agree with this. Many people can and do perceive the inexplainable issue directly. I had a discussion about that recently here on HN, about those who can perceive a "hard problem of consciousness" and those who cannot. Basically some people do not perceive any disparity between a totally physical model of the mind and our conscious subjective experience, while I and others personally do. It is a bit tiring to constantly be called delusional because I feel that disparity
Right but the idea that there are aspects of our minds that don't exist within this physical universe is not outlandish or bizarre, it's a very reasonable and historically defended position.
EDIT: Also not sure what this bizarre statement is:
> Mathematically, things with no evidence for their existence don't exist.
1. How is that related to mathematics? Can you provide a theorem or some definitions here?
2. How is evidence related to mathematics?
3. Why is mathematics relevant when talking about whether or not our mind has non-physical aspects?
>...the idea that there are aspects of our minds that don't exist within this physical universe is not outlandish or bizarre...
That's where we disagree. Can you give me an example of anything that doesn't exist within this physical universe? I'm not familiar with the concept of a "thing" that isn't physical.
Sorry about being confusing with the last sentence. I'll try and address your questions:
1. When we talk about the likelihood of something being true we are now discussing probabilities, which have a long and well proven track record of obeying the laws of statistics. Statistics are quite often counter-intuitive and our brains don't think in terms of the math that governs the actual outcomes
A good example is the birthday problem [1]. Imagine we had a room full of people and we wanted to know what the odds are that two of those people in the room had the same birthday. Obviously the more people we have, the more likely we are to have a match, but how many people do you need to get to that point?
It turns out that with just 23 people in the room there's a 50% chance of two people having the same birthday. Most people's intuition leads them to believe that for 50% you'd need 365/2 = 182.5 people, but once there's 70 people in a room, you're almost certain to have a match with a probability of 99.9%. This is a good example of a case where not knowing the math will lead your predictions of the world to be incorrect.
2. When we discuss evidence we are trying to figure out the probability of something, thus the rules of probabilities apply. One of the fundamental theorems of statistics is Bayes' Theorem [2], which at its core tells you how to calculate the probability of an event given another event.
To motivate that, here's a more realistic example: "A patient goes to see a doctor. The doctor performs a test with 99 percent reliability--that is, 99 percent of people who are sick test positive and 99 percent of the healthy people test negative. The doctor knows that only 1 percent of the people in the country are sick. Now the question is: if the patient tests positive, what are the chances the patient is sick?"
If you answered 99% then you'd not only be wrong, but you'd have misdiagnosed your patient. The correct answer here is 50% [3].
3. To bring it all together: to properly know things one must understand the laws of mathematics. This applies not just to problems that obviously involve probabilities, but everything in general.
At it's core, I think our disagreement revolves around this idea of non-physicality and the belief that there are things that aren't governed by mathematics. Everything is physical and obeys the laws of physics which themselves are governed by the laws of mathematics.
Is "mathematics" a thing in our physical universe? Many people argue it's an abstract set of constructions with no physical reality in themselves. Sounds kind of outlandish and bizarre according to your criteria.
> Everything is physical and obeys the laws of physics which themselves are governed by the laws of mathematics.
If you define "physical" to mean "literally everything" then of course having something outside of Physics is logically impossible. That is a different definition to the work "physical" that we are working with in terms of the mind. In that case "physical" doesn't mean "literally everything" but rather more like "those things that correspond to an external, verifiable, objective reality". For you, ideas and concepts would be considered physical, but that is not the case for the definition of "physical" I mentioned, which is far more common in both religion and philosophy.
In this case we would say that ideas and concepts are not physical. So an easy answer to your question "Can you give me an example of anything that doesn't exist within this physical universe? I'm not familiar with the concept of a "thing" that isn't physical." would be literally any concept or idea. However since you have a universalist (and thus totally useless) definition of "physical", of course it is impossible for me to argue that there is anything outside of that.
PS. As a doctor in quantum physics myself, I can tell you that statistics and probability is certainly a way of modelling outcomes, and not in any way considered a fundamental element of reality.
I'm afraid I still don't understand. Are you arguing that there is some utility in considering "concepts or ideas" non-physical from an ontological perspective? Because I think we both agree they're solidly physical in the sense that a "thought" can be reduced to chemical signaling.
On the QM side, I'm pretty sure that you're incorrect in your postscript. It was my understanding that certain physical processes can only be understood though probability and that talking in terms of wavefunctions is the most correct way to do that.
> Because I think we both agree they're solidly physical in the sense that a "thought" can be reduced to chemical signaling.
I don’t agree with that. I think there is an inseparable spiritual element that justifies the subjective experience.
For the QM part, that still does not mean that ontologically the probabilities have a physical reality. Google “interpretations of quantum mechanics”, we have very little idea of what quantum mechanics actually means yet. We can set up the problems and do the maths but no one even knows what exactly it means to do a measurement (again Google “measurement problem”). I am an expert in quantum mechanics (it is my job) so I would be happy to answer questions about it in the morning.
I think you're getting mixed up. Consciousness is obviously real (because we directly experience it) and is a hard problem because it's so weird.
The soul is not obviously real. There's zero evidence for it. It's just some random hypothesis that doesn't even explain anything. It's like reincarnation. Just a nice idea that someone made up with zero evidence, no mechanism and no use. Clearly nonsense.
How do you know it is nonsense? I don’t understand how you’ve categorically decided this. I’m a Buddhist so I don’t believe in a soul but I do believe in non physical aspects of the mind and also rebirth. It is actually possible to verify these things by Buddhist practise. You have decided it definitely isn’t true, but you haven’t got any proof either way. How is it then obviously “clearly nonsense”? That is just your preconception. If you are a scientist you should at least be agnostic, since we do not have evidence either way.
at first glance i thought this would be an opensource virus. which is a fun idea because as its updated it inspires people to patch their code. or an opensource virus that patches vulnerabilities is also fun.
You may be interested in meterpreter and the metasploit framework.
It's not an actual worm or virus, because that'd be a little silly, but from automated exploits and a good reverse shell that's a small epsilon away from what you're describing.
I imagine trademark lawyers would deal with that name pretty quickly. But, there are other cute animals which would look good blowing themselves up with bazookas, like hedgehogs.
Basically, while the structure of the network has been known for decades, they don't yet have the technology to figure out how the neurons behave (the weights of the neurons) nor how they learn (changes in the weights), so they can't "upload" data from a real worm.
I'm not an expert in anything relevant to this, so this is ill-informed speculation.
But isn't referring to not knowing "the weights" (as the article and researchers do) making a rather large assumption about the nature of the problem? It sounds more like the researchers don't know "why the neurons fire when they do", which would leave open the question of whether "weights" is even a sufficient model.
Oh I think it's an awesome project, too. It's just that it sort of makes concrete what were previously hypothetical arguments about simulating living beings.
I can't bring myself to care much about a C. elegans, but let's say we scale this up enough to simulate something the size of a dog plus its environment. Presumably it has subjective experience, like a real dog? What are the ethics of writing code that would result in its "suffering"?
What are the ethics of writing code that would result in its "suffering"?
Do you have the same concerns about other, perhaps much simpler but also virtual and pseudointelligent entities, such as computer-controlled players in games?
It does raise some deeply philosophical questions about life and what it means in general.
Nah, I'm not particularly worried about that. Because those are all currently based on models too simple to conceivably have anything like consciousness.
On the other hand, once we can simulate a human and its environment at OpenWorm's targeted level of abstraction, presumably we would have a conscious being there?
The only consciousness we experience is our own. In the extreme, we can no more say a chair suffers than a human being other than ourselves. Of course we take the pragmatic approach of deciding things that look like suffering to us is suffering.
I strongly disagree. The human brain, in its connectome, encompasses enough computational complexity to contain consciousness and suffering. A chair does not, at least not based on any model I've yet seen proposed.
Believe they meant the wood composing the chair came from a tree with complexity enough to be reasonably thought capable of consciousness (on some basic tree level)
Although a chair does subsume human projection, similar to a Buddhist realization of artifice.
A chair serves, regardless of its consciousness, but the human (or mammal stand-in) may project a consciousness of its utility upon the object, thus elevating its stature in a shared conscious.
The closest I can think is a friend establishing rules around an heirloom coffee table that enforced coasters and denied foot rest.
But I think that more organic models should have that assumption baked in to their more humanistic appropriation.
Like we kill a cow for sustenance, but I respect that in purchase by eating it.
A table serves me infinitely. And thus it carries the consciousness of the human.
Destruction of a utility reflects the human qualities, and so an inanimate object might adopt those qualities independent of time.
And on what basis do you claim that computational complexity is a necessary condition for consciousness and suffering? There is only one datum you have for consciousness. Other people can tell you that they experience consciousness, but so can a Text-to-Speech program
> Of course we take the pragmatic approach of deciding things that look like suffering to us is suffering.
Hence why I said this. There is nothing you can use to believe a human's claims over the text-to-speech's claims, so we choose what looks like suffering to us. However, my point is that it is an arbitrary choice.
> There is nothing you can use to believe a human's claims over the text-to-speech's claims
Of course there is. The human facing me looks similar to me so I can interpolate its claims with my own experience of being an human.
I thinks it’s also for the same reason that we have variable degrees of empathy against animals : the more the animal looks like us (that can be size, number of members, physically, in terms of effective communication…), the more, on average, we have empathy for them.
This can go to the extent that people commonly feel « something » about their cars and their human-face-like designs combined with their single ability to move.
Yes, but this is arbitrary. That is my point. There's no reason to believe that you can extrapolate your own subjective experience to others based on their similarity to you.
But you are choosing to use the Turing test, which tests a form of intelligence, as a proxy to determine whether to believe the claim. A Turing test does not preclude philosophical zombies. It does not demonstrate anything about consciousness.
"What are the ethics of writing code that would result in its "suffering"? "
I doubt that code can suffer. We do not understand feelings enough and this might help to understand it eventually - but for now I see no reason to assume that they can recreated digitally.
And I am much more worried, about many real life beings that suffer in our food chain and co.
Why would we assume subjective experience can't be recreated digitally?
Inert "code" can't suffer, of course. But assuming we accept the materialistic view, what reason do we have to believe that a mind in a simulated body has any less subjective experience than a mind in a "real" body?
"what reason do we have to believe that a mind in a simulated body has any less subjective experience than a mind in a "real" body? "
That reality is seemingly indefinitely complex and any simulation we have, very simplifying. I am pretty sure, for emotions and co, it takes a bit more sophisticated simulation - if it can be achieved digitally, at all.
Not entirely related, I wonder if they have simulations out there for the primordial soup and accelerating it (evolution) so that it hopefully gets to a human brain.
The primordial soup was presumably an Earth-scale massively parallel chemical "computation" with no goal. How would you even begin to simulate that? We at least know roughly what a worm does.
This is reeeeeaaaaalllllly hard. Remember that protein folding was unsolved five years ago. (I'm not an expert at all, and I can't make a ruling on whether AlphaFold is now generally considered a "solution".) Protein folding is just one small part of physics.
Yeah I'm not saying it's easy, I know I can't do it now haha. I'm a JS kiddie myself, still have yet to get into this kind of stuff like making a basic 3D simulation, best I got to is 3JS at this time. I'm working on it when I have time.
I watched some protein folding stuff from the Comma AI guy I didn't even know what was going on there but it looked cool.
But we do have computation readily available, my 3060 is just sitting there while I sleep, I could do some things with it maybe. Granted it's not a super computer, lower end.
I would instead call this "vastly underestimating the complexity of physics". Again, protein folding is a tiny tiny subset of physics and is famously extremely computationally difficult.
not only “not easy”, I highly doubt it is even theoretically possible in this universe. Probably it can be proved that a computer that simulates the universe must have at least as many qubits as atoms in the universe to be able to describe the state of the universe or something. Obviously not possible since qubits are physical entities
Don't mean to drag this on, not arguing with you guys but, can you not do a small part, like a puddle. Not the entire universe. I realize that's still a lot but you know, maybe you could do small sets/chunk it but yeah it's all bs on my part.
It is probably outside the human civilisation's current abilities to simulate the molecular physics of a human cell. Looking through the references of https://link.springer.com/chapter/10.1007/978-3-030-43229-4_... turns up in particular https://journals.sagepub.com/doi/10.1177/1094342018819741 from the year 2019, which mentions simulating "20 trillion atoms" on a certain supercomputer; this is perhaps enough atoms to make a very small cell, but I presume those simulated atoms are somehow very nicely behaved (and are simulated over a very short simulated timeframe) relative to the actual biology that happens in humans.
We can't even simulate the dynamics of a simple atomic system in an exact way, without making assumptions. The universe is that but with many many more atoms and complicated couplings. The universe is much more powerful of a simulation than any computer can ever be. Even if you used a quantum computer you'd need as many qubits as that original primordial soup.
Our computers are not more powerful than the "computer of the Universe", they are much _much_ less powerful.
https://www.lesswrong.com/posts/mHqQxwKuzZS69CXX5/whole-brai...
Apparently there are some signs that progress is being made again now, though...