An open source platform for emulating the fruit fly brain | Hacker News Reader

An open source platform for emulating the fruit fly brain

neurokernel.github.io

125 pointsmlla17 comments

This is really cool, also check out http://www.openworm.org/ which was started around the same time. I wish there was a comparison of the two projects.

These projects are amazing, and I only wish I knew about them a few years earlier. I studied C. elegans quite a bit when I was in school, and I also worked with D. melanogaster for a genetics project. I wonder how OpenWorm will help shape research in the years to come as it matures. It's a really nice surprise to see useful, open software that it so pertinent to the things I spent so much time studying - I hope I can still reach the people who picked up on my old projects and share these tools with them.

This is really cool. The other fruit fly project I was impressed by recently was Mihir Garimella's FlyBot: Mimicking Fruit Fly Response Patterns for Threat Evasion [1] which was a quadcopter that demonstrated evasive behavior. I really like looking at biomimicry implementations because there are so many levels on which to do it. Mimir was trying to mimic outwardly observable behavior while Neurokernel is trying to mimic neural pathways.

[1] https://www.googlesciencefair.com/projects/en/2014/6d1893c20...

Please note that while interesting in the context of a 14-year old executing it (under heavy supervision, I suspect based on the trajectory model), the translation from fly or bee to robot has been done for almost thirty years on a much larger scale. If you're interested, check out the work of Srinivasan, Borst, or this cool piece of engineering:

http://www.nature.com/nature/journal/v497/n7447/full/nature1...

This is very important. When people talk about emulating/uploading human brains, they extrapolate computing power and scanning to a point ~2040 when we will have the capability. But guess what, we have the capacity to upload a roundworm now or a fly and have tried to, and we cannot. Wy? it seems that there is something we are missing and most likely it's the science. tl;dr we need to understand brains better before we can upload a human or a fly, this kind of project will get us there.

I would be more impressed by your argument if these projects seemed to have any support or real energy behind them besides a handful of hobbyists and grad students. There's snake oil with far more funding and backing than these projects.

As far as I understand, the endgame of this stuff is a piece of software which would make a robot that was an exact physical equivalent of a fly without a brain behave exactly as a real fly does, assuming we had the computing power to run it in real time. Are these projects actually going for that, or do they have some more modest goal for what they expect to achieve, say, before 2020?

Well if they scale this up a bit - they can use it to fly planes.

"Brain in a Dish Flies Plane" http://news.discovery.com/tech/robotics/brain-dish-flies-pla...

This project is a threat to the Linux kernel. Matthew Garrett may become interested again in fruit flies and stop finding and fixing the moist weird bugs in the ACPI/PM implementation of Linux.

I think the term kernel is being used as gpu kernels/programs; those created with OpenCL, etc..; So its a proper naming scheme.

Maybe we should blame the GPU folks for switching from "shader " to "kernel" for general computing shaders instead?

The term is actually used in the sense of an OS kernel; just as Linux, Darwin, etc. provide services required by software applications written for those platforms, so too does NK aim to provide those services required to construct and execute models of the fly brain on GPUs.

Yes, i agree with you this can be a proper OS kernel too; i've misundestood the OP statement because i didnt know the real context of what hi was implying; my bad.

This is actually, probably a good step into what a "OS" kernel will look like in a not-far-away future

Argh, it is horrible when you have to explain jokes.

Matthew Garrett (mjg59 on HN) is a well known Linux developer. [1] He has a real talent for discovering and fixing weird, _weird_, problems with hardware firmwares, especially issues related to ACPI and UEFI. Have a look at his Linux.conf.au 2014 talk [2] to understand the kind of weirdness we are talking about (You will want to cry and laugh at the same time).

Before becoming a kernel hacker, Matthew Garrett earned a PhD in genetics studying fruit flies.

Now this project is very cool, both from the programming point of view and from the fruit fly point of view. Hence the fear that Garrett may abandon his spectacular work on the Linux kernel to go back to his original academic interests.

[1] http://en.wikipedia.org/wiki/Matthew_Garrett [2] https://www.youtube.com/watch?v=QP9c7aCZqtU

Oh thank you for the context; i've misunderstood what you have said than; i thought you were joking about a "fruit fly kernel" not being a proper kernel to replace the linux kernel, with some sarcasm added to it

Sorry, stupidity mine

Great project. After poking around the site, I'm still wondering, what GPU hardware is necessary to run the code?

Any Nvidia card made in the last 5 years will do.

Are we running this VM inside another VM?

Reply on news.ycombinator.com