Could you please explain why not? 548,000 synapses sounds entirely feasible to me.
I'm sure we're not far from making a high-level LLM-ish model of behavior based on those extensive studies.
But the topic of discussion is not that, but making a model sufficiently accurate that you "turn the crank" and it yields similar behavior without any priors of what that behavior should be.
To do that, at a minimum, we'd need for each neuron, the profile of responses to each of the neurotransmitters at each synapse, the excitatory/inhibitory effects of each signal, the patterns of how each neuron reacts to those inputs (i.e., receiving a signal from upstream neuron 489327 does not mean that it'll just pass it downstream, but that it'll decide depending on rate of firing, other current excitatory/inhibitory inputs, etc., if and at what rate it'll send the signal downstream), the rate of learning in each of those neurons... and a bunch of other variables, fully modeled.
Then, compute all of those running through the system, and have it take an input like a photo and output the same behavior, from the bottom up, without hints from the behavioral studies.
What am I missing here?
Even in the rest of your comment you’re taking a bit more of a bottom-up approach relative to what I’m saying: you’d be surprised how much we know about how the brain’s gross organization leads to complex phenomena (pick up the latest edition of Blumenfeld’s clinical neuroanatomy if you want the very-high-level summary).
You can, in principle, achieve a “broadly correct” outcome by doing tissue-level modelling of NNs. It’s surprising how much of the brain is macro components, as opposed to micro, cell-level processing. (Of course I’m handwaving a lot here. I’m afraid anything short of a concrete demonstration is bound to be unsatisfying.)
And yes, modeling at the higher functional level can be very useful; knock out the Wernicke's center and speech goes, visual cortex, vision, etc... So, with a more detailed functional description of each level, we may wind up with a model with useful predictive value.
Tho, that said, how does this approach create a truly robust abstraction from the lower level wetware? Would it provide ability to fully reproduce computations? Would it account for lower-level changes in health, hormones, electrolyte levels, neurotransmitter-active drugs...?
- If you model top-down (at the tissue/functional level like you suggest), you'll be recreating the "broadly correct" kind of computation, and you'll be recreating something that looks and feels like a really really bad, quirky and dumb human brain. But it'll have certain human-like qualities that are maybe even difficult to pin down. These would possibly make a ton of mistakes, but there'd be a lot of human-like biases and mistakes.
- If you throw billions of neurons into a bag, you may be able to train them to perform calculations with a high degree of correctness (eg.: ChatGPT, generative art, modern ADAS systems) but when these make mistakes, the mistakes they make will look extraordinarily stupid to a human (eg.: "a human would never have suddenly steered his car into a brick wall like that").
Both approaches can produce extremely stupid results, but you need the top-down architecture if you want to preserve what makes "the human flavour of intelligence" what it is. (I suppose you could emulate the same result with a big enough bag of neurons, but that sounds very inefficient to me, intuitively.)
---
Depending on what you're interested in modelling, you may need to combine multiple approaches, as the brain has multiple layers of emergent properties. I don't think that for most purposes you'd need to go as far as modelling blood contents, but something like it might be required if "embodiment" was an important part of what you'd like to model. There are certain types of things that biological organisms learn particularly fast because they have a physical body that interacts with the real world.
I don't personally believe embodiment is fundamentally required in the model (ie.: I think it's probably possible to emulate the same result if you use a sufficiently large number of neurons), but I think realistically it will be a practical necessity for keeping models and computations as efficient as possible.
"multiple layers of emergent properties."
I think that is the key, right there! And we need to choose which layers are necessary, sufficient, and/or useful for the purpose.
Change the expression levels of some genes, alter an amino acid here or there, change the input parameters (e.g. make a phenotype that lacks the ability to feel pain), and you can end up with a neuron network that responds subtly, or grossly, differently than another. Put memory into the picture and responses can be learned. Fruit fly behavior permanently changes in response to serious injury, similar to the manner in which humans experience chronic pain. Sure, now this is something that can be modeled, but before the experiment it wouldn't have been modeled.
https://www.sciencedaily.com/releases/2019/07/190712120244.h...
> After the injury healed, they found the fly's other legs had become hypersensitive. "After the animal is hurt once badly, they are hypersensitive and try to protect themselves for the rest of their lives," said Associate Professor Neely. "That's kind of cool and intuitive."
> "The fly is receiving 'pain' messages from its body that then go through sensory neurons to the ventral nerve cord, the fly's version of our spinal cord. In this nerve cord are inhibitory neurons that act like a 'gate' to allow or block pain perception based on the context," Associate Professor Neely said. "After the injury, the injured nerve dumps all its cargo in the nerve cord and kills all the brakes, forever. Then the rest of the animal doesn't have brakes on its 'pain'. The 'pain' threshold changes and now they are hypervigilant."
Are you saying this because you've spent time reading and/or researching fruit fly behavior studies? Or for some other reason?
I worked closely with one prof (not my PI) who specializes in the study of mitochondrial metabolism in fruit flies, but fruit flies are not my area of expertise. I know just enough to know that there’s an enormous amount of literature on the subject.
Yes, it's true that modelling just a single cell's interactions with its environments is beyond our capacity.
But here we're talking about simulating how a brain reacts to signals at a higher level of abstraction: we're studying an "emergent" phenomenon. We don't need to model molecular interactions, and we absolutely can model this using artificial neural networks.
This doesn't mean we'll be able to completely accurately model its behaviour, but we should get a lot closer than many HN commenters seem to believe. Biological neurons don't have magical properties, they just have more side-effects.
> we're studying an "emergent" phenomenon
It's like saying that mapping all cells in human organism to addresses in memory will give us emergent human inside computer.
But still at abstract layer we just do not have a model of a working brain, we only have maps of neurons, not an actual "algorithm" or a model of all relevant environment interactions needed for a working brain or a set of "abstract instructions" needed for it to work. I think you are over optimistic about current state of our knowledge if we can't even model organism with few thousand neurons accurately. Imagine trying to do that for tens of billions of neurons and trillions of synapses, especially that we know brain doesn't work like an ANN at all[1].
Sounds more like they're saying: a knee is a knee, it doesn't have any magical properties. Build something that bends, and it will behave like a knee. I don't know how true that is, obviously.
If you're active, knee implants have to be replaced after 15-20 years, because they detach from the bone. Knee implants can harbor pockets of infections because they don't have an immune system. They are also more prone to dislocations.
They are objectively inferior to a healthy organic knee.
But most of the reasons why artificial knees are not as good as the real deal have to do with the fact that they're made of inert (albeit fancy) materials. They don't have the ability to continually heal and do tissue remodelling, which is what real tissue does.
I feel very optimistic when I think about this: we're limited, but I think it's absolutely wonderful what we're able to do.
Hope that reality isn't 50-years or more into the distant future.
I'm sorry, I'm not quite sure how your profession is relevant to this discussion. Are you a biochemist who studies neurons and the ability to simulate them? That would be relevant.
I can think of at least 4 other broad specialties that would be more relevant to the discussion: neuroscientist, computational biologist, neurologist, computer scientist who specializes in simulations of biological systems, etc.
So was the first line just a reaching appeal to authority? If not, in my opinion you should include more context about how your experience is relevant in the future.
The actual precision of this model is: nobody knows, because nobody knows precisely what neurons do / what they react to. We know some of it but definitely not all. But, simulating what we do know, you get quite worm-like behavior, despite whatever flaws exist.
To get a more perfect simulation, we'd need more perfect knowledge of the chemistry and physics, and lots and lots more computing power. It's something that's continually improving, but a lot of shortcuts have to be made to make it even remotely calculable.
A fruit fly brain is a better model to simulate in this regard. It's much more "generic", so there's a hope that we can recapture high-level behavior from it more easily.
This is a completely baseless assumption. It is also moot, since simulating even 300 neurons is beyond our current computational power; simulating 3000 is not going to be possible in the foreseeable future.
(Doesn't sound like many?, but I've read that one single neuron needed 100 or so nodes in a deep neural net to simulate. But 300*100 also doesn't sound like many (nodes)?)
Edit: some explanations here: https://news.ycombinator.com/item?id=35113498 (in this discussion)
Massively simplified models exist, and yeah - we have artificial neutral networks FAR larger than that and they run just fine. But they're so over-simplified that it's fair to call them something else entirely, not a simulation. You can use them to create similar behavior, but they're just following the basic concept of a brain, not how they actually work.
Kinda like how a door hinge is not your knee, even though they both bend. It's not a knee simulation, it's just something that bends. For some things (doors) that's perfectly fine, and there are billions of them in use. You can't transparently replace your knee with one though, it has biology-juice all over it and needs to handle very different behaviors at times.
For brains: we have no idea. Every brain is somewhat different, but we only have a couple complete physical layout maps so far. No "weights", no specific chemistry per cell, no knowledge of what may have changed before it could be frozen for imaging.
With a lot of work and a lot of extra physical and chemical simulation (a full body in quite a lot of detail), based on that map and what we know of the rest of its body, openworm achieves wormy wriggling.
That's honestly pretty good! And it implies that the connection structure is at least coarsely meaningful in an extremely simple mind, which is what many have suspected but have had no way to verify before. But we know next to nothing about how that individual worm's brain behaved compared to the simulation. The fly brain will be similar - we'll have a bigger and more complex brain to run, more complex behaviors to compare against other flies, etc. It'll provide more evidence in favor of or opposed to the importance of that structure, compared to other things that we don't have enough information on yet.
Basically it's still extremely early days, so it's sorta like asking Stonehenge astronomers whether or not our supernova-brightness-based measurements of the size of the universe are accurate. They can see supernovas too, but their answer has to be "uh. maybe? give us a few thousand years to research it". We need more data and better tools than currently exist.
The neurons in worm directly interact with muscle cells, for example, via peptide/monoamine signalling. Or with "remote" neurons.
This is all possible because neural network of the worms is directly integrated with everything. And the connectivity is hard-coded in its genome.
It's not true for more complex organisms, simply because you can't encode that kind of complexity in DNA. Instead it encodes the overall connectivity structure, so that emergent behavior provides necessary instincts.
Just as there's more than one way of generating electricity there may be many ways of generating thought.
However we definitely have the computational ability to do simulations a fly network. Look at some of the modeling done by the Blue Brain Project or Allen Institute for Brain Science - they do simulations of rat and mice models with hundreds-of-thousands to millions of neurons and exponentially more synapses. 3000 neurons is not that many. If you stuck to non-compartmental point models a 3,000 neuron simulation could probably be ran on a moderately high-end laptop.
But as said before, the physical connectome is only part of the information you'd need do any worthwhile simulations.
It's not the complete picture though, normally that brain would be in an ever-changing soup of chemicals, which definitely impact behavior... somehow. Simulating that, and even knowing what might be relevant to simulate, will never be complete. Only incrementally better than previous attempts.
Trying to simulate a 3000 cells and 500K connections of the fly brain is not a computational problem, it's a knowledge one. If you can find functional properties to build a spiking/rates model, and data to compare it too; then it would be feasible (although a lot of work) to build and run simulations on the model. But without that extra info, and only using the physical connectome, there would be very little reason to try to do so.
Investing the search space, to invent a name, can sometimes be very powerful.
No way.
First, this mapping doesn't tell us how the synapses are regulated - if we could 'run' this the weights would stay fixed forever and that's not how a brain works.
Second, there must be some neurons dedicated to chemical management, and they'd go haywire unless you found a way to deal with them. It's possible the hardware/software is so intertwined it can't be separated. Or maybe it's just complex, regardless the mapping is of limited use here.
Third, you're assuming that the synapses/axons are the only thing that matters. It's may well be this is true, but having other processes being involved has not yet been entirely ruled out. If they are, the mapping is incomplete.
Lastly, we don't have the computational ability just yet to simulate even the mapping itself.