Neuronal wiring diagram of an adult brain | Hacker News Reader

Neuronal wiring diagram of an adult brain

biorxiv.org

80 pointsmpweiher39 comments

The lead author of the paper, Sven Dorkenwald, shared more resources on Twitter. Here are excerpts:

"""

We are releasing a whole-brain connectome of the fruit fly, including ~130k annotated neurons and tens of millions of typed synapses!

Explore the connectome: https://codex.flywire.ai

Reconstruction paper: https://biorxiv.org/content/10.1101/2023.06.27.546656v1

Annotation paper: https://biorxiv.org/content/10.1101/2023.06.27.546055v1

There are multiple ways to access the data and the number of tools is growing! https://flywire.ai/apps

Codex: https://codex.flywire.ai

CATMAID spaces: https://fafb-flywire.catmaid.org http://Braincircuits.io: https://braincircuits.io/app?p=fruitfly_fafb_flywire_public

"""

- Original Tweet Thread: https://twitter.com/sdorkenw/status/1674859033076072448

How unfortunate they chose to post to a closed platform.

Twitter users may as well not exist to the rest of the world.

Only a matter of time before this goes the way of the open worm project. Incredible that they’ve increased the number of neurons imageable by this magnitude. A similar increase from here would encompass the human brain. Models of the human brain would likely be indistinguishable in every way, shape, and form from the real thing.

> Models of the human brain would likely be indistinguishable in every way, shape, and form from the real thing.

Eventually, but the more I learn, the more I think that's going to be an ethical nightmare to even test.

First brain to get the We Are Bob treatment is going to get almost every neurotransmitter slightly wrong, and be tripping as if they're suddenly alternating between ketamine/cocaine, LSD/antipsychotics, and every other pair of opposite impact substances you can think of.

I keep hearing doctors saying things like "we don't really know how general anaesthetic works" and "we don't know why sleep and dreams are biologically necessary, only what happens if you miss them" for example, and those feel like important things to understand when building models for simulation.

The neuron models used in the open worm project… function…

That is to say that the complete organism model can learn to swim, which about all the organism is supposed to be able to do.

I see what you’re saying though, that human-level function will likely be much more difficult to fine-tune.

The ethics of it all are a philosophical wormhole. Ultimately, its “consciousness” would be an image of a real person. I don’t think we would have to worry about it too much. Saying it was truly conscious would be like saying that a video of a person was conscious.

> Saying it was truly conscious would be like saying that a video of a person was conscious.

I hope we believe what's true, regardless of what that happens to be — we get bad futures either way, if we think sims are conscious when they're not, or if we think they're not when they are.

I've used the film/TV analogy myself, just over a year ago, when it was GPT-3 (and LaMDA behind an NDA) rather than ChatGPT et al: https://kitsunesoftware.wordpress.com/2022/06/18/lamda-turin...

A video you can interact with and has its own thoughts, feelings, motivations and such; sure.

> going to be an ethical nightmare

Thanks for the We Are Legion / We Are Bob reference - that one wasn't on my radar.

My point of reference for the sort of horror that mind-uploading and simulation might result in is this fantastic short story: https://qntm.org/mmacevedo

As pointed out by other comments, this paper is about an adult fly, so we're not there yet, but still...

What will we do with a completed connectome?

Imagine handing a late 1800s scholar a modern processor. Give him a wiring diagram, and a voltmeter. What could he learn from it?

The connectome will be a big step, but not the penultimate one, in solving human intelligence.

Maybe trying to reverse engineer it using GNNs until they produce the same output(within some error margin) for various arbitrary inputs? Dunno.

I am sure there’s a great deal we can learn about how the brain actually functions from such simulations.

However GNNs don’t really capture the underlying mechanisms very well. Our brains use different kinds of Neurons which also use many different neural transmitters, their location in 3D space matters, as do stuff like blood alcohol content, O2 levels, adrenaline, signal propagation delays, etc it’s a stupidly complicated system even above and beyond the number of neurons.

Also limiting it to neurons ignores all of the activity and modulatuon that's increasingly being attributed to astrocytes and other CNS "scaffolding"

As many things as possible would have to be taken into account. Maybe a different mathematical formalization is necessary.

"penultimate" means "next-to-last"; I'm thinking you used the wrong word, or are making an obscure point I don't follow.

>Models of the human brain would likely be indistinguishable in every way, shape, and form from the real thing.

Except all those pesky qualia...

Can you distinguish between a system that did and a system that didn't have qualia? I think that would be a major breakthrough.

In the correct physics engine, with the correct models of eyeballs etc., a connectome-based model may experience qualia. It would likely tell us that it experienced qualia.

The biggest difference would be that new connections wouldn’t be formed, so it wouldn’t learn.

>In the correct physics engine, with the correct models of eyeballs etc., a connectome-based model may experience qualia. It would likely tell us that it experienced qualia.

Sure. It would say it touched water. It would just not actually be wet. Which is something...

When I dream of swimming, do I have the qualia of wetness at the time, or does that only happen when I wake and remember the dream?

Any neuroscientists here: does anyone know that much about dreams? My memory of philosophy lessons was that philosophy doesn't keep up with the latest in real science…

but the corresponding cluster of neurons in a brain isn't actually wetted by the touched water either.

Yes, but it receives the input from an actual experience of the body it's embodied in - from an actual wet thing. As opposed to being served a simulation of the same.

This simulated isolated "brain" is like a heart mechanically assisted to be pounding in a glass jar. Sure, it's still beating, but not as part of its place and environment, in a living being.

Where would you say experience ends, and qualia begins? The way these words are being used seems to make little distinction between them other than one referring to a specific instance of what the other refers to in generalized terms.

As for myself I would argue that an experience, and therefore qualia, amount to an accumulated and generalized succession of multiple sensory inputs amalgamated by feedback and hysteresis over time, and informed by the gross structure and constraints of the sensory apparatus they are carried upon.

Not so different a question than asking where “synapses and neurotransmitters” end and the experience of consciousness begins.

Models of the human brain would likely be indistinguishable in every way, shape, and form from the real thing.

Models that are just a map of connections certainly wouldn't indistinguishable from a human brain. Every single one of the trillions of human neurons is a quite complicated object. Neurons communicate with each other in a number of ways and neuroscience isn't even certain that communication methods have been found.

This discussion is in the context of the open worm project. It has successfully modeled the neurons of a nematode such that the organism model functions inside a physics engine as would the real organism.

Extending this to humans, the neuron models are good enough. I’ve done enough biomedical engineering research in the neuroscience domain to tell you: the problem is imaging the connections, not modeling the neurons.

Interestingly, the NeuroMechFly project [1] already exists, so integrating the two together may already be under way. At 50 million synapses, a modest gaming PC should be able to run the simulation. I wonder what people will think once they are able to do this at home?

[1] https://prelights.biologists.com/highlights/neuromechfly-a-n...

Holy misleading title, batman. The linked preprint is talking about an adult _fruit fly_ brain:

>> Here, we present the first neuronal wiring diagram of a whole adult brain, containing 5x10^7 chemical synapses between ~130,000 neurons reconstructed from a female Drosophila melanogaster.

Really need some clarification, that title does.

In the article they say openness is really important. But when you to the website, you have to use Google Account... It's open yeah, but someone wants to track you on this...

Being tracked is different, and does not negate openly available. You don't get into every library in the world simply by turning up: identity has been a part of access since writing was invented. You can't request books from many libraries without identifying yourself and the works requested.

In most libraries in the US, browsing the collection--even taking books from the shelves to a table for a while to read and make notes--does not require any sort of identification whatsoever. It's only when we temporarily deprive the library of a copy of a work, i.e., borrow it, that we have to identify ourselves.

In public libraries sure, but not in many academic libraries or other private ones. Which is of course where the more interesting (not mass-produced) content is held.

And nobody would call those open.

That's not quite the same. I'm not required to make an account with Google to go to the library. Being registered by a government agency and being tracked by multi-national that is famous for abusing tracking data are on a different planet imo.

The problem, for me at least, is that I do not use google products. I do not have an account and do not want one. So for me, this is not openly available. We could argue about the merits of being so principled in not using google, but that seems besides the point :) I should be free to no do business with Google and if that precludes me from accessing something, it is not openly available I'd say

On the contrary in every library I have been to you just walk in and browse the books. Last time I went to a university library a librarian was more than willing to help me find the books I needed and didn't ask any questions at all regarding my identity. I dare say that she would have refused to let me walk out with the books but that's understandable and anyway I wanted to refer to them not take them away.

I was able to legally download the entire research PDF directly from the linked-to-URL — without logging in [and I do not have any Alphabet accounts, with DNS-block].

I reckon I didn't say it: I was looking at downloading the datasets, not the PDF.

Thank you for this feedback. I am on the FlyWire Codex team, and we are working to 1) provide alternate login means and 2) open up parts of Codex to browsing without login

Why so salty? It clearly says additional identification methods are coming, and that if you don't have a Google account you can email to get help signing in.

Also of interest is the paper, “A leaky integrate-and-fire computational model based on the connectome of the entire adult Drosophila brain reveals insights into sensorimotor processing”, which discusses the results of simulations of this dataset with the Brian2 spiking neural net simulator. They activate the same input neurons in the simulator and real flies and show that there is similar downstream behavior in the sim and real neural circuits.

Reply on news.ycombinator.com