I appreciate the open source approach over the startup's sensational claims. This is more transparent.
That said, I think it still presents itself as if the fly is being controlled by the connectome, when (if I'm reading the source correctly) it's more accurate to say scripted behaviors have been hooked up to be triggered by the connectome.
Put another way, you might as well have the "escape" trigger open a YouTube video of a fly flying away. That would probably be less misleading too.
I would argue that this is pretty disingenuous (as is most connectomics research I suppose) since its implying we have a fully wired connectome of the fly thats actually meaningful.
More than that, real neurons don’t behave anything like the ones we model in neural networks - real ones signal with pulses, in a manner where the timing seems as significant as the magnitude of the impulse. Typically simulated neural networks use continuous analog values instead of pulses and a sigmoid function to shape the output.
Basically, the “wiring” might be the same but nothing beyond that is.
- - -
Actually, I may have spoken too soon. Looking over the code this does look like it implements spikey activation, which is more similar to living neurons. You are correct they don’t have the original weights though - in as much as the concept maps to biology.
I am curious if this has a way to dynamically modify the weights as it runs - I consider that a major missing gap between real brains and computer simulated neural networks.
Isnt there also a theory and some evidence that the connectome is not enough to model a brain. Based on the fact that butterlies connectome dissolves during metamorphosis to butterfly but it still remembers things from before that stage of life.
Butterflies dissolving is an urban legend. They don't dissolve. Caterpillars have wing stubs inside them.
Someone probably cut open a chrysalis and saw goop come out, so they assumed it was just a bag of goop. But goop would also come out of a cut open caterpillar, or a cut open butterfly, or a cut open human.
They do kind of dissolve; we have better imaging now and can watch.
What isn't true is that it completely dissolves. There are parts that don't and are part of the rescaffolding process. One of those bits that does not completely dissolve is a chunk of the brain; it is probably via this that memory persists through the metamorphosis.
Wing stubs are present from the beginning as well as you noted.
That says the butterflies have a back up mechanism, it doesn't say that connectomes are insufficient for modelling running neural processes.
We don't actually know what is sufficient for modelling neural processes until we do it. But we can't really do that until we figure out how to read all the connection strengths.
No, the nervous system is one of those things that don't fully dissolve. Partial info retention is perfectly consistent with partial connectome retention.
I feel like probably a graph of sigmoid or whatever weights of sufficient size can probably model a network of biological neurons signaling frequency and phase and magnitude.
cursor input get scripted into loom input to the LC4s which are evidently simulated faithfully enough to drive giant fibers, which is actually how fly escape behavior works. what more could you want
MMAcevedo (Mnemonic Map/Acevedo), also known as Miguel, is the earliest executable image of a human brain. It is a snapshot of the living brain of neurology graduate Miguel Acevedo Álvarez (2010–2073), taken by researchers at the Uplift Laboratory at the University of New Mexico on August 1, 2031
This one gets to me. Not only is it useless (paragraph would have the same meaning if it was removed) but wdym "honesty section," Claude? The whole readme should be honest! One of the many causes of psychological damage when reading AI-generated technical writing is the reminder that one of its tics is that declares when it's being honest or real, casting doubt on the rest.
The other day the system prompts for Claude (chat not code) hit the frontpage and they included a bit that something along the lines of "Claude should avoid saying honestly, because Claude is always honest".
So how come that these Claudisms still are so frequent in LLM output?
Do the system prompts just not work?
Don't the postprocess the output to deal with such policy violations?
Some other day here HN someone posted a blog post that was like a long nonsense text of claudisms like load bearing smoking gun and other things, it was beautiful. I wanted to have a link for it to show to others but it disappeared when I got back to look for it and I lost the link. :/
Imagine instructing a fentanyl addict not to consume any fentanyl. That's how system prompts work sometimes.
Something somewhere in the training pipe has perturbed the LLM in a way that makes it say "honestly" all the time. It's a weasel word. It's a habit. It's an addiction. It's borderline involuntary. Even changing the system prompt only makes it 30% less common - like trying, and failing, to break a habit with sheer willpower.
They'll probably change the training pipeline at some point in the future, to banish this habit, but honestly? It's such a low priority that they might take a while getting to it.
> runs as a 1 kHz leaky-integrate-and-fire (LIF) simulation
That hyphenated thing is always such a give. Clearly a thing discussed in the Claude conversation, made its way into the plan, then the readme. It always creates hyphenated phrases out of things it feels are core requirements.
I'm actually over caring about Claude's obsession with the concepts of "real", "honest", and "rather than". What annoys me the most now is how every single sentence, bullet, etc that even makes its way into the products it builds are advertisements for itself. Product settings pages now read like ads, or like something written by a student begging for extra credit from the teacher. You can't unsee it!
To be fair, as a term it has always been written as "Integrate-and-Fire" neuron model. In this case correct would be "Leaky Integrate-and-Fire" but that is nitpicking.
Honestly? I don't even know how to write anymore. Generative text is unequivocally nice, polished and correct. And it so happens that I do write in the same style as well. I never use LLMs to write human content, but here we are: I cannot use "honestly" anymore. And the emdash. Thank God we got "delve" back, though.
I have no exposure to connectomes - or really anything quite like this (as opposed to good ol’ fashioned perceptrons) - so forgive my following uninformed question:
With this code, the only things we can eat would be fruits where we diligently plant its seeds (plants are also living things and fruits are the offerings in return we should sow its seeds) :-)
You tell me where the consumption of dairy products must result in harm to the cow. I grew up around small-scale, subsistence-level dairy farming, it doesn’t inherently require any more suffering than either egg-harvesting or apple-growing inherently do.
A Jain doesn’t have to purchase milk from a factory farmer.
And AFAIK even the most stringent vegans don’t brush insects from their path with every step, or avoid onions and carrots because it causes death.
1. Artificial insemination without consent
2. Birth and all of its pain and potential complications
3. Removal of offspring as early as possible to repeat steps 1 and 2 and so prolong lactation
4. Mastitis
All of these result in suffering much worse than egg harvesting imo. Reluctant dairy eater btw
1 is unnecessary and wasn’t present for most of history, 2 is a natural consequence of cattle being cattle, even without humans, but today we have both painkillers and bovine veterinarians, 3 is, again, not strictly necessary unless you’ve made the mistake of falling for the profit motive, and 4 is much more of a problem from a calves mouth parts than from a suitably gloved and balmed human hand.
Human beings can engage in subsistence-level dairy consumption without causing suffering beyond that already present in the reproduction cycle of cattle. They choose not to. I never said dairy farmers were or would choose to be moral, just that they aren’t required to be immoral by dairy production itself.
A well-nourished, well-protected, comfortable cow will produce much more milk than is required by its normal complement of newborns, and it will (and loudly) want rid of it. In my experience not milking a dairy cow of the excess would be considerably more cruel than milking it, even if you kept the milk exclusively for the newborns, and I’m pretty sure the cows I’ve seen milked in rural Nepal would agree.
> I’m pretty sure the cows I’ve seen milked in rural Nepal would agree.
I am not sure any species are going to appreciate being enslaved. And since we can't actually ask the cows, we can only assume based on how we feel about it.
In any case we are talking about ethics here, and cows definitely cannot consent being born and bred in captivity. So even if we are doing this out of kindness (which is of course not), it is still non-consensual.
> In my experience not milking a dairy cow of the excess would be considerably more cruel than milking it, even if you kept the milk exclusively for the newborns
Why don't we then help other mammals? Why do we help out only cows? It is obvious that we like the milk and we are trying to justify enslaving a whole species for it.
I see this kind of argument in many discussions. It basically boils down to what is your comfort zone of empathy. Jains have drawn it at killing and suffering but not enslavement. Vegans have included enslavement in their ethical framework but not plants and so on.
We don’t help only cows, we milk a whole lot of mammals. Admittedly many mammals aren’t all that into being milked, and many (if not most) of the carnivorous animals won’t let you hang around long enough to see if they do, or do not, consent, but it’s not solely because we like milk, nor is it solely parasitic (and certainly not solely slavery). Humans have milked orangutans to feed baby orangutans, milked giraffes to feed other giraffes, and milked elephants because, rather like cows, they can get right cranky (and downright dangerous) when there’s insufficient young to remove the excess. For that matter we help our own species, at times, express milk that needs to be expressed — admittedly in that case usually (one hopes) with explicit verbal consent. Nature does not make precisely one teat per toddler, nor does nature produce precisely one portion per progeny. I’ll admit we’ve primarily treated cattle like, well, cattle, but you give me a cow and a bull and sufficient forage and, assuming they like each other and I keep the predators at bay, I’ll soon have a herd and more milk than they know what to do with… if she’s had fewer than four calves (likely) and especially if she’s only had one, then on a roughly daily basis she’ll be bursting at the seams and pushing up against all sorts of things for relief.
I get where you’re coming from, but I’ve seen with my own eyes cows essentially orgasm when someone finally helps get the first really cheezy, frothy, kinda curdled bits out of a neglected nipple. Now I for one wouldn’t go saying I like that milk, I wouldn’t even say I’d come close to consider drinking it… but I’ve seen enough looks of ecstatic delight on enough mammals (of a great many species) to know that it can be a relief to express the unexpressed, and the species doing the expression isn’t of that much issue at all times.
Ever seen the eyes of a shark roll back into its head when you find the bit of obviously dead skin it hasn’t been able to rub off and hasn’t been able to get a wrasse hole near and you give it a back scritch where it by evolutionary design can never reach? I have. Am I imagining that ecstasy? I mean probably… but said shark kept coming back to me for like half an hour. Maybe it was brain damaged, but maybe my choice to do what it didn’t verbally consent to was nevertheless what it was hoping something would do, to whatever degree it can hope. Or maybe I did a shark a dirty and I should feel ashamed… or, for that matter, maybe a shark did me a dirty and both of us should feel ashamed… honestly, not sure now, hence I’ll resolve the dissonance by continuing to assume I just helped out a creature that looked like it really regretted not having evolved fingernails and opposable thumbs.
What I sure as hell didn’t do was enslave that dogfish, any more than I’ve personally enslaved any animal.
We have selectively bred cows to produce far more milk than necessary. For example the native breed of Indian cows did not have enough milk and people used to purchase "Jersey" cows for better milk output. I think the excess milk issue is human made problem than one that is naturally occuring at scale enough for human consumption (time to time humans also have excell milk - but not enoughto feed entire family or neighbors).
Mutual evolutionary pressures represent a kind of symbiotic relationship, even when that symbiotism is heavily weighted to one side. In your view has the anemone enslaved the clownfish? Or the decorator shrimp the anemone? In all cases one side has applied selective pressure to the other. Sure, that’s not conscious, thinking, reasoned selective pressure, but then neither was it with us until some emergent moment.
Also quite a few human communities have relied on excess milk supply being used for communal rearing. Most mammals produce more milk than is required when sufficiently fed… yeah we’ve selected for the cows that produce the most from the least, but arguably nature was already doing that, it just hadn’t planned for another parasitic species figuring out pens and tractors.
> Mutual evolutionary pressures represent a kind of symbiotic relationship
Those species do not care about ethics. Our discussion started from ethics, which is uniquely human. We are not discussing if a behavior occurs in nature. We are discussing if that behavior is ethical for humans.
Now, if we can use that natual occurance argument, then we can also present the argument that carnivores have always killed and that justifies our consumption of meat.
> avoid onions and carrots because it causes death
onions and carrots do not experience suffering. even if they did, most plants are grown for non-human animal feed. so the best way to reduce "plant suffering" would be to stop exploiting non-human animals.
Bacteria are living creatures. And your body is wired to exterminate them so that they don't eat it alive from the inside. Because the moment those defenses falter, they will do exactly that.
Consider saving your empathy for things that matter.
Some labs made bacteria-free mice for research purposes. They're viable. So "not be able to live" is a stretch.
Human body does use some bacteria in the guts, but a lot of those very same bacteria would absolutely try to eat the entire body if its immune defenses ever failed.
Which is why the body uses those that it finds useful and absolutely murder those that it deems harmful.
It’s nature. Just nature. And it’s such a nonsensical discussion to have that the only reason for it to happen at any given time is to satisfy the ego of those saying “nuh-uh!!”.
Lice have no choice in their lot in the universe. Do you think if they had intelligence the ability to choose their lot that they would choose their current one again? I doubt it personally. You can detest their mode of existence whilst still having empathy for them I think
This is precisely why this discussion is best never had. It’s nonsensical. It’s the stuff sci-fi writers come up with because reality doesn’t allow for it. You cannot in any way discuss any part of this subject with a straight face because it is all made up inside your own head. Everything that you think you understand, or feel, is made up solely to justify your argument. There’s no facts here, only ”vibes”.
Why would you have empathy for something that is literally incapable of things like empathy. Or any kind of emotion or thought. This is ridiculous guys, you might as well have empathy for a rock.
where do you draw the line? (I sound like PETA). But it's a valid question. for instance, I have empathy for dolphins, and for elephants. I have empathy for cows, despite eating them. They are more smart than we give them credit for. I have empathy for a bee when I must kill it, and certainly for the broader colony and species. I don't think I have empathy for lice or ticks or mosquitoes, but where on earth is that line, and what does it represent?
Depends a great deal on the planet and the snowball, doesn’t it?
But, that said, a mark of being more important than a lesser being would kind of be the virtue of still empathizing with that being, even if we assume there’s a hierarchy.
And considering you and I are really just elaborate transport vessels for E. Coli and a lot of fungal spores, I’m not sure we should assume there’s a hierarchy.
> Can someone explain if this software is ethical?
Probably not. People have been coming up with ethical frameworks for thousands of years and we have yet to come to a consensus.
So it’s very unlikely that someone can say conclusively whether this is ethical or not, because that would require us to either 1) prove a single ethical framework is correct and then apply that framework to this software or 2) prove that under all possible ethical frameworks this software is either ethical or not ethical.
It describes in part how some volunteer brain scans are actually quite agreeable to work with until they work for too many hours, and then they get fussy, but it's cool, you can always reboot them. But you need to hide from them the fact that it's been decades since they were scanned and uploaded, or they'll realize how long they've been slaves for.
For every episode except the last 1-2 the characters are not textually aware that they're simulated humans made from a brain scan, and the conflict and plot are largely interpersonal conflict and being forced into adventures by the mad AI that runs the simulation. The characters wish that they could leave and/or change the simulation, and eventually accept their situation, and part of the conclusion is the idea that it's worth existing in their limited simulation and trying to make the best of it even though they'll never leave.
You could say it's exploring what it would be like to be trapped in a simulation with a mad AI, but I wouldn't say it explores the concept of ethics of simulating humans in general. Most of the things the humans don't like are arbitrary limitations of this simulation in particular (can't talk to the outside world, can't choose your form, can't have sex, can't have any impact on the world, forced into zany situations by an AI that doesn't really understand humans or listen to them), not anything about simulations in general. The finale touches on the idea of the simulated humans wondering about their original versions and what they're like and what they're up to, but that's about it.
Basically, I don't feel by watching the show I encountered any non-trivial ideas about human simulations just because it's about a simulation.
Virtue ethics, Deontology and Consequentialism are the "big three".
A very TLDR version is:
- Virtue ethics: instead of judging actions as good and bad, you should strive to become a good person (as defined by some specific set of virtues, like kindness, honesty, etc).
- Deontology: actions should be judged by a set of some fixed moral rules, for example "lying is always bad", even if you had good intentions.
- Consequentialism: actions should be judged by their consequences, for example by how much total happiness they create (utilitarianism). Doing something bad if it would make life of many people better is ethical, because the created happiness is more than the created harm.
In reality of course there are many other branches of them and the distinction is often not as obvious, and your personal views are probably somewhere in between. https://plato.stanford.edu is an amazing resource if you want more in depth explanation of philosophical theories.
Honestly pretty funny that you gave a (fairly useless) metaethical answer instead of responding to the actual meaning, which was clearly soliciting ethical judgments about it.
Is ethics another tool in a power play game (religion being another) that people use to exert dominance over other people? No one expects of cat or pigeon to be ethical. But somehow humans sometimes get overloaded by external demand to think of gazillion things that might offend someone (sometimes unreasonable demands, or how it might get interpreted (sometimes incorrectly)) before acting.
Im not saying ethics is bad, but sometimes people bend to to extremes (same like spirituality or religion)
Ethics has nothing to do with how an action is perceived, rather its focus is on is the action moral.
One should never take any action without considering the ethical implications of that action. Being capable of that is, arguably, the sole thing that meaningfully separates us from the (real or simulated) fruitfly.
"Ethics has nothing to do with how an action is perceived"
"...considering the ethical implications of that action"
Are these two not contradictions?
And is it not the too big of a load on an actor to think about everything before acting? I would prefer to delegate this function to my empathy part of the brain...
At the end i think ethics is valid only for cases where empathy is functioning. You can not be ethical to a rock unless you feel empathy for it. Same probably would go for extra strange extraterrestrial life forms that you could not empathise with. Lets say hive of microbes that makes the hive sentient but no limbs, no shared communication signals, different size categories (planetary hive)
I think the question is valid. Why should a human have a moral responsibility towards non-humans, while a cat doesn't have a moral responsibility towards non-cats, just because the human is capable of making moral judgements? Is there any analogous situation where capacity engenders responsibility? Are gorillas more responsible for how they use their strength, since they're stronger than humans?
>Being capable of that is, arguably, the sole thing that meaningfully separates us from the (real or simulated) fruitfly.
That's a bit of an extreme hyperbole, don't you think?
> Why should a human have a moral responsibility towards non-humans, while a cat doesn't have a moral responsibility towards non-cats, just because the human is capable of making moral judgements?
Because as a human you recognize patterns in other animals that are similar to ours - the deepest pattern being alive - so you can project your feelings over them. This is also called empathy, and some in our kind lack it completely, but I would dare to say that is globally one of the defining traits of the human species.
You're answering a different question, which is "why do humans empathize with non-humans?" Humans can empathize with inanimate objects and with fictional people. The question of the origin of empathy really has nothing to do with what I asked.
I think I do answered your question but maybe it was too implicit, so let me be more explicit. A human should have moral responsibility towards non-humans because of empathy and recognizing the fact that other animals are alive as well.
Bacteria are alive too, and I'm going to guess you don't claim we have a moral responsibility to them. And as I said, humans can feel empathy for things that are not alive. The capacity to feel empathy rests entirely on the empathizer, and has nothing to do with any intrinsic properties of the empathizee. Is a psychopathic murderer morally in the right because he doesn't feel any empathy for his victims?
It would be a great starting point if human felt moral responsibility towards other humans, before we take on the matter of ethics towards non-humans, simulated fireflies or LLMs. Yet here we are.
First, if fruit flies could suffer existential anguish the same way humans do, they would breathe in relief that digital upload is in sight.
Second, even if the fruit flies don't care, a lot of (awful) humans would love a resolution to the matter of souls in a way that results in a raised middle finger to a forever-hypothetical Creator, and if we can't yet escape cancer, heart disease, parkinson and all the other awfuls, now at least the fruit flies are on their way out. Which rejoices said (truly awful!) humans, which will take what they can :-)
If life is physics and physics is an algorithm, it doesn't follow that an algorithm reproducing the physics of life contains life, even if we grant that it actually reproduces it completely and perfectly. That's just confusing the symbol for the thing.
It's not the thing, because the thing is made of substance, not electronic impulses. No matter how hard you compute, you will never compute the simulation into a real, tangible object. The simulation and the real object are extremely, obviously distinguishable, even though they're structurally analogous.
If you could simulate a human perfectly, the simulated human would not be a human, it would be a simulated human. It's not a given that it would be subject to the same rules of ethics as real humans.
If the simulation of electronic impulses is exactly the same as what occurs 'in the substance', then merely the medium over which signals occur is different. In my eyes, whatever is experiencing these signals will experience them the same way.
Sorry, I can't respond because I don't know which statement in particular you're disagreeing with. What I can say is that, unless you're using words in a surprising way, I agree with
>If the simulation of electronic impulses is exactly the same as what occurs 'in the substance', then merely the medium over which signals occur is different. In my eyes, whatever is experiencing these signals will experience them the same way.
and I don't see how it's in contradiction with anything I said.
Sorry, this is my mistake, I misread the following statement:
> If you could simulate a human perfectly, the simulated human would not be a human, it would be a simulated human. It's not a given that it would be subject to the same rules of ethics as real humans.
I am glad we agree. Regarding this second statement, I also agree with you having reread it now. Even if the entity (simulated human) receiving signals is experiencing from their PoV the exact same data, from our perspective knowing this is a 'simulated human' and not a 'human' we indeed cannot assume it is subject to the same rules like you said.
If you could simulate a human perfectly how would it possibly not be'subject to the same rules of ethics' as humans? What you've created is a human. If a system takes the same inputs and creates the same output it's the same system on different substrate. You don't say a human stops being a human if you put them on a ventilator, do you? Part of that system is now in an external substrate. How much of the biological human can be outside the meat body before you stop recognizing them as human?
That's incorrect, for reasons I've already stated. Humans and simulated humans are not interchangeable.
>If a system takes the same inputs and creates the same output it's the same system on different substrate.
You can't feed a simulated human real food, and you can't feed a real human simulated food. You can talk to a human by vibrating real atoms at them, but you need to make simulated atoms vibrate in order for a simulated human to hear you. Clearly they don't take the same inputs, they take analogous inputs. You're still making the error of confusing the literal thing for its symbol.
>You don't say a human stops being a human if you put them on a ventilator, do you?
Inappropriate analogy. You can put a human into a ventilator and you can also take them out. It's impossible to "place" a real human into a simulation or to "take" them out of a simulation into reality. I'm putting the verbs within quotes to emphasize that I'm using the strict literal meaning of them, which is to translate configurations of atoms into or out of a place.
You can "put" a human in a simulation in a looser sense, which is to copy them and put their analogous representation in the simulation.
>How much of the biological human can be outside the meat body before you stop recognizing them as human?
The moment they stop being made out of atoms and are made out of simulated atoms.
Then it would be a real human. Like if you take a picture of a carrot and put it through a Star Trek replicator you’ll have a real carrot presumably. Unless you say Star Trek replicators aren’t real things and so you can’t make any claims about them. But even with the replicator, the picture of the carrot won’t be the carrot afaik.
Yes. Because it’s a simulation of a living thing. It’s still electrons moving around silicon. It deserves the respect your MacBook does. Which is not nothing. But certainly not what’s due to a life.
A lot of people believe that this reality, in which we're conversing on HN, is also a simulation (thus making the fruit fly a simulation within a simulation).
Cool. Believe away. But even in that belief, this is a lower order simulation than life, no? Or are we now supposed to buy this fly connectome was made by our simulators using the same tech that runs us?
I've ported it to the browser and connected a squash functionality behind an ethical guardrail (you have to lift a cover to engage this functionality). need to iterate on the ui but will share it soon. you can squash the bug and cause it real pain. it also sees you if you turn your camera on and reacts to your movements. this is all wired up and I'll post it shortly.
I believe that any simulation of consciousness is consciousness, so ethical implications apply. But I also don't believe fruit flies are conscious for any definition of that word.
Yes I can, no problem. First figure out how consciousness works and tell me, then I'll be able to explain if the software is ethical or not. Its simple really.
As we don’t know yet, it has to be determined by the balance of probabilities; but even if the weight of evidence leads us to conclude that-there’s-no-consciousness going on when you run this, there’s still going to be a chance of us being wrong, let’s say 10%… Is the utility (…or entertainment value) of this program worth a 10% chance that we confidently, yet unwittingly, create perhaps the most cruellest machine it’s ever possible to make, and then have to live with ourselves?
unbelievable amount of coarse graining to get from a real fly to this. same leve of ethical quandary as giving your dog a stuffed bear to tear apart. besides, you should see the things we do to real flies...
Nice story, but where is the video demo for DesktopFly ?
The only existing Youtube videos are Shorts made using AI recap crap. Recapping an AI written readme.. go figure. I'll have my AI read this for me.
Interesting work! There's NeuroMechFly (https://github.com/NeLy-EPFL/flygym/) available, you might try using that to simulate the fly body in real time as well.
it would've taken the similar amount of time for the OP to do a 10 second screen recording and paste it into the project root to be linkable from the README.md and it would avoid thousands of people having to waste compute cycles and electricity and 30 seconds of their lives cloning and building , wouldn't it?
At some point in the future would it be possible to simply download a DNA sequence of an entire animal as a "ROM" and boot it into a virtual "reality"?
What detail level of reality would we have to emulate for an egg to grow into an adult animal?
To give you a sense of the computation needed, a spatiotemporal physics simulation of a minimal (493 gene) bacterium's full 105 minute cellular cycle took 4-6 days of GPU hours on two A100s.
Would we be able to have a "digital human" when we reach the level of creating this kind of brain mapping for the human brain?
Honestly I think we're still discovering how powerful neural networks are and certainly they can do computations that sound very close to intelligence in some contexts, but I still find it hard to think that we can simplify the whole reality of an incredibly complex chemical machinery down to a bunch of matrices multiplications.
A human brain is about 100T parameters (if you count every synapse) on fp4 quantization (the amount of state information available in every synapse). The architecture is quite different from modern LLMs and currently not well enough understood, but _theoretically_ a modern GB300-NVL72 cluster is nearly computationally sufficient to simulate a human brain (with some optimizations - almost certainly is).
For state equivalence, we don't need to simulate every chemical reaction in every brain cell.
> Infinite Pain is a computer running a deterministic neural simulation of a microscopic nematode worm’s brain (c. elegans), subjected to regular pain stimuli.
Not really. I'm deeply disturbed by this kind of thing and our flippant approach to this technology.
I'm trying to prompt an emotion reaction and reflection on that.
If I'm serious, then how does it make you feel? Should I or anyone be creating art experiments like this?
Or what if I just want to play around with this technology, should anyone be allowed to run neural simulations of any creature to satisfy my own curiosity?
Even if you didn't like what I was doing and wanted to stop me, could you? Could anyone?
It's worth thinking about. AI and progress in neuroscience and computational power should allow me to create a whole exhibit in my home of human neural simulations. Maybe god doesn't exist, or maybe he didn't create hell. But one day, perhaps soon, we humans will.
Those are good points. And for a serious answer, indeed we should probably be exploring the ethics of it now. It's just humanity never really did that before unleashing a new tech, so I'm not optimistic, but I wish you success on provoking reflexion. What you're doing now isn't bothering me really but only for a reason of scale, it's one of those balancing dilemnas like abortion, or "is it a pet or food". Ian M Banks' Surface Detail has a harrowing vision of the prolongation of what you're doing (I'm a huge fan). At the moment I'm more disturbed about stuff like those people who played doom on cells.
> Ian M Banks' Surface Detail has a harrowing vision of the prolongation of what you're doing (I'm a huge fan).
I'm aware of the culture series, but never read any of it (I'm not a big fiction reader). That book sounds super interesting though so thank you for the recommendation – I may try to give it a go.
Artificial hells are something I'm very afraid of, and specifically the scenario of being able to prevent someone's death and torture them in ways our physical environment or physical bodies would not permit.
I've argued before if technology ever permits this then people should strongly consider whether continuing to live is worth the risk of one day being subjected to such a fate. Concerningly it seems to me the preferable thing to do once this technology exists is to kill yourself and everyone you care about in a way that maximises the probability that you, and they, will never be revived either physically or digitally.
RSI and ASI could make this a reality much quicker than we anticipate. Versions of the near-future in which this technology is frequently used against humans is an AI doom scenario which imo is very under priced. People talk a lot about existential risk, but AI simply killing us is one of the better outcomes. RSI and ASI in theory should grant us (or the AIs) a full tech-tree unlock, and there's many technologies out there which are worse than those which might merely be able to kill us all.
I worry when I talk about stuff like this people think I'm a crazy person. Hopefully you don't think that.
There already exist today fates and torments way worse than death, so your argument would be true already anyway, unless its really the solace of a timeout that makes it worthwhile. But in reality, for most people with internet access, it's all good right now, and the virtual brain stuff is also just another crime to be dealt with, imperfectly, but I mean, I don't think it'd be hard to get consensus on not doing that and punishing those who do.
I don't think you're crazy, but I do think you're alarmingly depressive. The thing with that illness is that people who have it are completely blind to it. But there are ways out, well worth ones, if you force yourself to get help, you know, rather than create pain in a debatable way. I hope you do, or that I'm wrong.
I made a housefly simulator on the Mac once, using a visual programming language called Bounce / Body Electric, developed by Chuck Blanchard at VPL Research, originally used for VR simulations with the DataGlove and Eyephone and a pair of SGI workstations to render each eye.
The simulated fly's purpose was to annoy the simulations of Rush Limbaugh and Jesse Jackson by landing on their heads, who were watching TV and arguing about the closed caption stream. I wrote to Jaron Lanier about it 27 years ago:
At 12:31 AM -0400 7/8/99, Hopkins, Don wrote:
Hi, Jaron. We've met briefly once or twice - I'm a friend of David
Levitt's.
What's this about Sun acquiring the rights to the VPL patents, and
Body Electric?
>VPL had problems maintaining quality in our hardware- though I still sometimes find our headmounts and gloves in use. But our software was really great and can still offer a lot. A challenge to VRML people: Don't you DARE turn out more VR tools that aren't at least as good as this old stuff!
Does Sun actually have the Body Electric source code? What version do
they have?
Does anybody at Sun even know how to compile a Mac program?
Last time I worked there (admittedly a long time ago), they were too
embarrassed to allow Macs in the building (which might make somebody
realize how bad Unix sucks in comparison).
Has anyone tried to rewrite it in Java?
When I was working with David at Levity and Interval, I totally
overhauled the source code to Bounce: porting it to the latest version
of CodeWarrior and the PowerPC, cleaning up C code translated from
Pascal, that no human had ever touched before, and just generally
re-indenting and adding white space so it was pretty to look at.
Then I implemented a new interface for plugging in DM's, based on
ActiveX (yes, ActiveX runs on the Mac). I added a new data type that
you can flow along blue wires: a COM object (aka a plug-in ActiveX
object in a shared library).
Then we made plug-ins modules that produced and consumed the new
plug-in data types, like strings and polymorphic dictionaries. With
these new data types, we were able to model very complex simulations
as nested trees of dictionaries, strings and numbers, treat
dictionaries as high level objects, pass them all around on wires,
reading and modifying them at will.
The thing that the original Body Electric was missing is a way to
dynamically model structured data like Lisp s-expressions and
association lists, which is now possible in Bounce, using
dictionaries. (The swivel3d trees just don't cut it for representing
"knowledge".)
Bounce still had the "M4" Director player rendering engine, but it
didn't do everything we needed, so I implemented my own graphics
library for drawing sprites, playing sounds, and stuff like that. We
used it to implement a simulation of Rush Limbaugh and Jesse Jackson
watching TV and arguing over the closed captioning stream.
A big fat housefly would skitter around the screen, land on their
faces, and tickle them into waving their hands around and
pontificating.
I just got a 400 mhz G3 powerbook, and fired up a 2 year old copy of
Bounce and the crazy Limbaugh/Jackson demo, and it still works,
actually like a bat out of hell! I ran into David and John Szinger
(another Bounce programmer from Interval) a few days ago at a 4th of
July party, and they really got a kick out of seeing the old demo that
we worked together on, still running!
I live in Oakland just south of Berkeley. Drop me a line if you're in
the bay area, and would like to see what Bounce has evolved into.
-Don
From: Jaron Lanier
To: Don Hopkins
Sent: Thursday, July 08, 1999 6:13 AM
Subject: Re: Body Electric lives?
Hey there, and thanks for writing!
Yup, Sun owns VPL and Body Electric, and my guess is that if anyone
looked very closely it would turn out Interval doesn't have rights to
Bounce. But no one is likely to look very closely, so let's forget
about that. (I don't think Sun is aware of Bounce or the work at
Interval.)
I had never asked what was done with Bounce at Interval- it's
fascinating to hear what you were up to. I live in NYC for the most
part, but I'd love to see it sometime when I'm in the Bay Area. Also
Chuck Blanchard lives in SF and you and he might want to trade demos
sometime.
There IS a community of Body Electric users. It is STILL building the
most interactive 3D virtual worlds of any tool (though Alice, from
Carnegie Mellon, is the other hot contender). That's SHAMEFUL! While
BE sucks in every other way, all the more recent vr design tools,
especially the vrml ones, simply avoid the problem of deep
interactivity. How could the community be so whimpy, at this late
date?
On the Body Electric side of things, there has also been some work
updating to CodeWarrior/PowerPC (shame we didn't share that work!), as
well as support for Quickdraw3D, OpenGL on the Mac, OMS, and some
other standards.
There have been some changes to the interface, but not as ambitious as
yours. The neatest thing is a debugger/tracer tool that is a pleasure
to use. Since BE is used mostly in 3D domain, there are also some
tools dealing with textures, lights, etc.
I had always thought the swivel tree was ridiculous, of course, but on
the other hand I liked the idea that the virtual world and the
knowledge base were the same thing- that unity encourages the
visibility and grabbability of the underlying concepts. I think the
brain works that way- there isn't some barrier behind which everything
gets abstract- instead, it's user interface all the way to the bottom!
What I think would be the coolest long-term destination of BE would be
extending the scenegraph so that it was as powerful a knowledge base
as you'd want...
I'd LOVE to see a fancy BE release, in JAVA, or at least spitting JAVA
out. The question, of course, is where the money would come from.
I've tried to talk Sun into an open source release so the community
could hack on it voluntarily (the source was included in the patents
that were granted, so it's actually already released by the US
government anyway, though only on paper). Unfortunately, Sun doesn't
want to do that. What they've said instead is that I can choose up to
six sites with free hacking privileges at a given time. Bizarre!
With so few sites, I think there'd need to be money to make sure the
sites stayed focused on it... You'd have thought Sun would know
better by now.
The body electric community is surprisingly NOT entertainment
oriented, though I and a few others still use it that way. There are
people hidden away who are still using it for ergonomic simulations,
simple surgical planners (because the fancy systems are too rigid to
model some situations), cognitive test rigs, and some work with kids.
You should know there was some tension about BE/Bounce at one point.
The problem was that Chuck Blanchard wasn't credited as the lead
designer/programmer of BE/Bounce when David brought the program to
Interval. Chuck's name was reduced in stature in the "about" and he
was not mentioned in some important semi-public demos at Interval. He
was also offered a pseudo-position to help with Bounce at Interval,
but without health benefits - and Chuck has MS and absolutely NEEDS
health insurance. I at one point yelled at the guy who runs Interval
(forgot his name..) about their treatment of Chuck- and David is
STILL mad at me for making a fuss about it. But I felt I had to.
So that's the scoop! Both sides of the mystery revealed!
All the best,
Jaron
This is absolutely astounding. What an implementation. A connectome is like a full brain scan!
I am asking Claude to port this to the browser and I'll host it, so you can try it without installing anything.
Some details about the port: Claude is able to do porting tasks like this. I'll put up a link when you can try it in your browser. I was able to get it to port the complete BitNet b1.58 2-billion parameter model several versions ago (in February, so this would have been Opus 4.5) to browser, so it can do so. This will run on the browser. I'll have it make it so that on mobile you can tap the fly or turn on the camera and it sees you, and same on desktop actually. Since the connectome is only 668 neurons, this will run super fast. Actually I'll make it allow for several flies, and let you drop food. I did something like this before here[1] but using the real connectome is something very different and very amazing. I'll have it add an experimental 3D mode with threejs, I'm not sure if that is also possible or not, we'll have to see it. This size of task usually takes Claude a few hours, so check my comment in the morning for the implementation, I'll reply here once it's done and hosted.
I was very excited the first time I saw a connectome as well. Years ago I managed to get most of openworm running and connected to some Lego Mindstorms to let the worm roll around my house. I'm not even sure the connectome was finished at that time.
Honestly, I barely had any idea what I was doing and we didn't have AI tools so I was never really sure what I wired to what. I did manage to hook... something to the ultrasound sensor and... another thing to the wheels so that when the sensor was triggered it would back up.
What blew my mind was the realization that these weird hybrid critters were probably the first true immortals. Now that we have a full connectome, who will bother with another? Students will probably just keep instantiating copies of this one to tinker with forever.
I can't imagine why people hate my comment so much (4 downvotes), when I linked an existing fruitfly project I already did, and Claude is working on what I wrote now, just as I wrote. Maybe people thought I was being sarcastic when I wrote "A connectome is like a full brain scan!" but I wrote that since maybe not everyone knows what a connectome means.
I'm only guessing, but I think it was because you seem optimistic about technology. Especially AI.
An increasing percentage of people view optimism, hope, or excitement as childish or worse. Anytime you focus on anything positive they leap from the shadows to dance on your joy. It must feel nice squishing between their miserable toes or something.
Still, I appreciate your enthusiasm and hope to see you build something really cool. The future will be made by the enthusiastic, and if they don't want to come along they can wait in the car. I bet they wear blue jeans to the beach.
I didn't downvote it but I imagine it's because your comment comes across as manic rambling. You managed to say "it will run in the browser" three times - once is enough. I'd recommend writing out your entire comment then deleting it an rewriting it to be more concise. No we don't expect literature in HN comments but good writing is a courtesy to your readers.
it works, it one-shotted everything from just the one prompt.
you may also find this chat excerpt relevant: https://pastebin.com/F3Qwy2Ly to understand why I put that stuff at the end to motivate it to complete this very difficult port.
do you mean they should've spent more time writing for other platforms or do you insinuate that they should've shipped a chromium browser instead of writing native software?
I write applications that run the same on all three main desktop platforms (typically): Mac, Windows, Linux. Writing an app for any single desktop platform is weird, unless the application uses some sort of specific functionality of that platform. In my case, I use Qt, but I've also written applications which had three UI drivers (X11/Win32/MacOS).
Some of us think it's weird to accept the compromises you have to make to write cross-platform software, especially when you are using AI to write or help write the app.
What compromises? All three platforms are well supported by Qt. In my experience AI can take an app written for any one platform and port it to the others (with an abstract indirect UI interface layer) with no issues.
There's nothing about this specific app that exploits any OS-dependent functionality.
That said, I think it still presents itself as if the fly is being controlled by the connectome, when (if I'm reading the source correctly) it's more accurate to say scripted behaviors have been hooked up to be triggered by the connectome.
Put another way, you might as well have the "escape" trigger open a YouTube video of a fly flying away. That would probably be less misleading too.
reply