Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Science

Fly Brain Breakthrough 'Huge Leap' To Unlock Human Mind (bbc.com) 68

fjo3 shares a report from the BBC: They can walk, hover and the males can even sing love songs to woo mates -- all this with a brain that's tinier than a pinhead. Now for the first time scientists researching the brain of a fly have identified the position, shape and connections of every single one of its 130,000 cells and 50 million connections. It's the most detailed analysis of the brain of an adult animal ever produced. One leading brain specialist independent of the new research described the breakthrough as a "huge leap" in our understanding of our own brains. One of the research leaders said it would shed new light into the mechanism of thought." [...]

The images the scientists have produced, which have been published in the journal Nature, show a tangle of wiring that is as beautiful as it is complex. Its shape and structure holds the key to explaining how such a tiny organ can carry out so many powerful computational tasks. Developing a computer the size of a poppy seed capable of all these tasks is way beyond the ability of modern science. Dr Mala Murthy, another of the project's co-leaders, from Princeton University, said the new wiring diagram, known scientifically as a connectome, would be "transformative for neuroscientists." [...] The researchers have been able to identify separate circuits for many individual functions and show how they are connected. The wires involved with movement for example are at the base of the brain, whereas those for processing vision are towards the side. There are many more neurons involved in the latter because seeing requires much more computational power. While scientists already knew about the separate circuits they did not know how they were connected together.
Anyone can view and download the fly connectome here.
This discussion has been archived. No new comments can be posted.

Fly Brain Breakthrough 'Huge Leap' To Unlock Human Mind

Comments Filter:
  • It wasn't science fiction; that movie was predictive.

    • "Help me" (Score:4, Informative)

      by MrKaos ( 858439 ) on Thursday October 03, 2024 @10:11PM (#64838567) Journal

      It wasn't science fiction; that movie was predictive.

      it was also a remake.

  • check your work (Score:4, Interesting)

    by spaceman375 ( 780812 ) on Thursday October 03, 2024 @09:12PM (#64838497)
    The obvious next step is to emulate this connectome in a virtual fly and see how it behaves. I'm willing to bet it won't even be able to move properly. The chemicals that energize the connections are many and varied, and this counts. They need to map beyond organelle right down to localized chemical gradients before it will approximate a living fly.
    • True enough, but it is still progress. If you want to simulate a brain, you need to go to a much simpler creature. Let's start with a planaria and work up from there.
      • Re:check your work (Score:5, Interesting)

        by ShanghaiBill ( 739463 ) on Thursday October 03, 2024 @09:47PM (#64838531)

        If you want to simulate a brain, you need to go to a much simpler creature. Let's start with a planaria and work up from there.

        We've already started with nematodes, specifically C. elegans. Caenorhabditis elegans [wikipedia.org]

        The entire genome has been mapped, and some behaviors can be simulated. And it's all open source. OpenWorm [wikipedia.org].

        The cool thing about C. elegans is that all worms of the same gender (there are three genders) have the same number of neurons and identical connectomes. So, you can pick apart the brains of multiple worms and combine the results.

        Disclaimer: My daughter is a biologist. Worm brains are a normal dinner table conversation topic for my family.

        • by vbdasc ( 146051 )

            Worm brains are a normal dinner table conversation topic for my family.

          Hope they're not also the main dish on the dinner table.

    • We know where the neurons are, we still don't know how they work. We know how a lot of it works.
    • Re:check your work (Score:5, Interesting)

      by codebase7 ( 9682010 ) on Thursday October 03, 2024 @09:59PM (#64838551)

      The obvious next step is to emulate this connectome

      NOPE.

      Anyone with experience in emulation will tell you that trying to emulate the system at bus level logic requires exponentially more computational power than the real hardware itself has. Each one of these neurons is effectively it's own processor. (Able to process input signals and generate output.) Operating independently of all of the others. 130,000 of them. Even the best CPUs on the market don't have that kind of core count, let alone with their own independent cache. A supercomputer might get you there, but it will be very slow, and that's with high-level emulation where common functions and code are substituted with precompiled instructions in the host system's native instruction set. Which we cannot do here.

      Nevermind the need to accurately simulate the timing between neuron firings and getting "data" (simulated chemical signals) to the right places. Which isn't uniform due to different timing requirements for construction of the various chemical compounds and the differing amounts and positions of available raw materials to construct them. (Don't forget variance between individual neurons!) A modern computer really doesn't have a step like this, as the needed electrical levels are maintained passively. (Yes, the system can turn things on and off. It's a broad over simplification for the sake of those reading this.)

      TL;DR: We don't have anywhere near the artificial processing power to do this for a fly, let alone anything bigger.

      One other thing, even when emulating bus logic you still need the raw data and firmware or that massive undertaking isn't going to be doing much once it's fired up. We only have the hardware connections. We still need to dump a copy of Fly BIOS and OS, and figure out where that data goes physically, before we can do anything useful with it.

      • That's a very long & detailed way of repeating the sentence from TFS, "Developing a computer the size of a poppy seed capable of all these tasks is way beyond the ability of modern science."
      • You'd be mad to do this on a single machine, you'd need a cluster of them, or even a cloud!
        • Or just accept that it's going to run very slowly. Which is probably exactly what you want if your goal is to study how it works.
      • by smbell ( 974184 )

        You're right, if the goal was to emulate a fly in real time. I don't think we're anywhere near that stage.

        We absolutely could do this emulation, and I suspect that will be on the table soon. Yes it would be slow. How slow would depend on how much hardware we wanted to throw at the problem. We're not limited to a single computer, even a single supercomputer. The only thing really standing in the way is money.

        BTW, the BIOS and OS are the mapped neurons and connections. There's not something more to it.

        • A shocking number of people are heavily invested in either the belief that there is a mystical soul-like component that is completely beyond mortal knowing, or a mystical quantum blah blah sciency component that is totally beyond meatspace understanding. Then there are nuts that pick up one of those plus the bogus AI singularity theory, like doing what a fly does in the space of a poppy seed requires ridiculously complex iterations of present day technology but if we make a machine smart enough it's all dow

      • Anyone with experience in emulation will tell you that trying to emulate the system at bus level logic requires exponentially more computational power than the real hardware itself has. Each one of these neurons is effectively it's own processor. (Able to process input signals and generate output.) Operating independently of all of the others. 130,000 of them. Even the best CPUs on the market don't have that kind of core count, let alone with their own independent cache. A supercomputer might get you there,

      • by Plugh ( 27537 )
        An open an active question is the degree of fidelity to which we need to model individual neurons, in order to usefully model high-order cognitive activity. Spiking Neural Nets are more biologically realistic. But there are complex models of just 2-3 neurons that you need a big-ass cluster to run, because they model down to the level of cellular organelles and membrane chemistry. But from an engineering perspective it is easier to get high cognitive achievement modeling a billion simple "neurons" that each
      • You have received some neagtive or cynical comments, but I enjoyed your post, and I think you are right on the money.

        To add to that:

        - Bio-neurological structures operate at a "clock rate" that would test the patience of Rip van Winkle - milliseconds for neuron and axon response, microseconds for some of the chemistry that juices up the neurons, but that is still three orders of magnitude under digital electronic clock rates, even more compared to the best analog circuits. Computers can massage a spread she

      • 50,000,000 connections ÷ 130,000 cells = 384.6 connections per cell. Not too bad when you consider that the average human brain contains 100 trillion connections ÷ 100 billions cells = 1,000 connections per cell.

        Pretty shitty when you then compare "size of a pin head" to the size of the organ between our ears!
    • by Bongo ( 13261 )

      As a lay person, I hope you're right. The sense that life processes are way more complex than we seem to imagine.

      Also, mapping neurones, if that was it, then it isn't like there's such a thing as transferring information via fields. Nope. No such thing. Oops, why isn't my radio working...

    • Congratulations, you can read! "Developing a computer the size of a poppy seed capable of all these tasks is way beyond the ability of modern science."

      Actually, for /., that's pretty good reading comprehension ;)
      • "Developing a computer the size of a poppy seed with certain capabilities is beyond modern science" and "Simulating a brain the size of a poppy seed takes more computer power than a room-sized supercomputer" are completely orthogonal statements. We can build things we can't simulate, and we can simulate things we can't build.

  • by Rosco P. Coltrane ( 209368 ) on Thursday October 03, 2024 @09:27PM (#64838505)

    "Fly Brain Breakthrough"... For a second, I thought it was a RFK Jr. story and it wasn't just worms.

  • by phantomfive ( 622387 ) on Thursday October 03, 2024 @09:38PM (#64838517) Journal
    It took over a decade to get the first map of a brain, and that one only had 300 neurons. The brain mapping technology is improving dramatically. https://singularityhub.com/201... [singularityhub.com]
  • This sort of research represents the first steps in the direction that AI research needs to go if we're ever going to achieve real full-blown AI.
  • A fly's brain is far more advanced than a human's, how can it possibly have similarity to dumb-ass human brains?

  • One can study an FM radio for decades, but if one believes the entire time that the radio device is producing the content, one will never grasp the reality of the situation.

    Likewise, one could study a laptop for decades, but if one believes the entire time that the google search results and all the websites are spontaneously generated within the laptop, one will never grasp reality.

    Is conciousness generated within matter? Or does it pre-exist and animate matter? Both are possibilities.

    Today's scientists

    • Nice examples. Figuring out the hardware is step one. Understanding the software (wetware?) is step 2. But it clearly doesn't end there.
    • Obviously you're hinting towards "souls" using brains as a proxy. I'm guessing this interesting theory conveniently aligns with your spiritual beliefs. Good for you I guess.
      Do you have an issue with humans creating synthetic brains that these "souls" can use as a proxy? Do your ancient religious texts forbid such a thing? Don't your want to set free as many of these dormant souls as possible? Imagine server farms full of them, perhaps living together in some form of virtual "matrix" (for want of a better w
    • by fuzzyfuzzyfungus ( 1223518 ) on Friday October 04, 2024 @06:58AM (#64839053) Journal
      That's utter nonsense: if one studies a radio in any detail it's going to become abundantly clear that basically everything except the audio amplifier and UI components is focused around interpreting the electrical current emerging from an otherwise passive piece of metal. If your model of physics does not include electromagnetic fields for some reason investigation of how exactly those currents emerge from the antenna may hit a wall; but even then you'll be able to determine that that's the area of interest, that radios with antennas removed do not produce output, radios with antennas replaced by direct electrical stimulation can again be made to produce output, etc.

      The laptop case would be considerably more complex; but you'd run into the same ability to chase certain high level behavior down to interactions with the NIC driver; which itself interacts with a particular device on the system bus; and even if, for whatever reason, the idea of data transmission outside the chassis is off the table(which would be pretty weird when inside the chassis it's pretty much nothing but components chatting with one another; for for sake of argument); you'd still be able to observe that the behavior that is of interest has particular dependencies that you can break in various ways to observe changes in high level behavior; or stub out and replace with substitutes that emulate their behavior to some degree, like pointing a browser at a local webserver or constructing a custom peripheral that emulates the NIC on the PCIe side in order to inject inputs into the driver.

      It's true that you won't understand how the system actually works if you are specifically incapable of grasping some essential principle of it; but what you can do is determine(potentially in considerable detail) what the mystery behavior must be capable of, and where it must act upon the system, in order for what you've mapped out to have behavior that matches what you are seeing. If it's the case that 'consciousness' is animating matter, say, a sufficiently detailed teardown of something that is conscious should reveal deviations in certain areas from apparently similar electrical or chemical behavior in things that aren't. Perhaps it will reveal nothing about how that 'animation' occurs; but not all ignorance is equal: someone who can't figure out why results are 100mv off what the biochemistry says they should be in certain synapses is like an astronomer who can confirm that observations of Mercury's orbit deviate are not quite in accordance with what newtonian mechanics would predict. In the worst case they may remain in a state where their best answer is a shrug or "uh, maybe magic pixies?" rather than general relativity; but their ability to define their area of ignorance will actually be pretty precise.
  • There is a hug gap between mapping and understanding how it works. This is not a "huge leap". The announcement is hype for more funding
    • Well, there's a lot of money for funding AI projects sloshing around ATM. A tiny fraction of that could go a long way to better understanding what it is that all these startups & corporations claim they're trying to do... oh... erm... or... Maybe exposing the sheer difficulty & likely time-scale of doing what their claiming would make investors baulk? The singularity is only days away, right?
    • There is a hug gap between mapping and understanding how it works. This is not a "huge leap". The announcement is hype for more funding

      Mapping seems like a prerequisite for understanding, though, and the difference between not having a map and having a map is a huge leap. You can take a huge leap forward and yet still be very far from your intended destination.

  • This is a good opportunity to pass on a recommendation I think I got from slashdot originally a decade or more ago.

    The book "To know a fly" by Vincent Gaston Dethier is a great short read, I read it to my son as a bedtime book.

  • Read an article suggesting that only the largest pathways/nodes were mapped. Might a reference to the resolution permitted by the technology. However, if there are smaller features then this map is far from complete. If Penrose [wikipedia.org] is correct then distribution of smaller features would be just as important.

Many people are unenthusiastic about their work.

Working...