Neuralink looks to the public to solve a seemingly impossible problem
Neuralink looks to the public to solve a seemingly impossible problem
Elon Musk's quest to wirelessly connect human brains with machines has run into a seemingly impossible obstacle, experts say. The company is now asking the public for help finding a solution.
Musk's startup Neuralink, which is in the early stages of testing in human subjects, is pitched as a brain implant that will let people control computers and other devices using their thoughts. Some of Musk's predictions for the technology include letting paralyzed people "walk again and use their arms normally."
Turning brain signals into computer inputs means transmitting a lot of data very quickly. A problem for Neuralink is that the implant generates about 200 times more brain data per second than it can currently wirelessly transmit. Now, the company is seeking a new algorithm that can transmit this data in a smaller package — a process called compression — through a public challenge.
As a barebones web page announcing the Neuralink Compression Challenge posted on Thursday explains, "[greater than] 200x compression is needed." The winning solution must also run in real time, and at low power.
Submit your algorithms under GPL
AGPL just in case they try to put your brain waves into the cloud
GPLv3, make it really radioactive to them.
I’m not an Information Theory guy, but I am aware that, regardless of how clever one might hope to be, there is a theoretical limit on how compressed any given set of information could possibly be; and this is particularly true for the lossless compression demanded by this challenge.
Quote from the article:
The skepticism is well-founded, said Karl Martin, chief technology officer of data science company Integrate.ai. Martin's PhD thesis at the University of Toronto focused on data compression and security.
Neuralink's brainwave signals are compressible at ratios of around 2 to 1 and up to 7 to 1, he said in an email. But 200 to 1 "is far beyond what we expect to be the fundamental limit of possibility."
The implication of a 200 to 1 algorithm would be that the data they're collecting is almost entirely noise. Specifically that 99.5% of all the data is noise. In theory if they had sufficient processing in the implant they could filter the data down before transmission thus reducing the bandwidth usage by 99.5%. It seems like it would be fairly trivial to prove that any such 200 to 1 compression algorithm would be indistinguishable in function from a noise filter on the raw data.
It's not quite the same situation, but this should show some of the issues with this: https://matt.might.net/articles/why-infinite-or-guaranteed-file-compression-is-impossible/
Absolutely, they need a better filter and on-board processing. It is like they are just gathering and transmitting for external processing instead of cherry picking the data matching an action that is previously trained and sending it as an output.
I'm guessing they kept the processing power low because of heat or power availability, they wanted to have that quiet "sleek" puck instead of a brick with a fanned heatsink. Maybe they should consider a jaunty hat to hide the hardware.
Gathering all the data available has future utility, but their data transmission bottleneck makes that capability to gather data worthless. They are trying to leap way too far ahead with too high of a vanity prioritization and getting bit for it, about par for the course with an Elon project.
There is a way they could make the majority of it noise - if they reduced their expectations to only picking up a single type of signal, like thinking of pressing a red button, and tossing anything that doesn't roughly match that signal. But then they wouldn't have their super fancy futuristic human-robot mind meld dream, or dream of introducing a dystopian nightmare where the government can read your thoughts...
I'm not sure that's accurate.
Take video for example. Using different algorithms you can get a video down half the file size of the original. But with another algorithm you can get it down to 1/4 another can get it down to 1/10. If appropriate quality settings are used, the highly compressed video can look just as good as the original. The algorithm isn't getting rid of noise, it's finding better ways to express the data. Generally the fancier the algorithm, the more tricks it's using, the smaller you can get the data, but it's also usually harder to unpack.
I'm no expert in this subject either, but a theoretical limit could be beyond 200x - depending on the data.
For example, a basic compression approach is to use a lookup table that allows you to map large values to smaller lookup ids. So, if the possible data only contains 2 values: One consisting of 10,000 letter 'a's. The other is 10,000 letter 'b's. We can map the first to number 1 and the second to number 2. With this lookup in place, a compressed value of "12211" would uncompress to 50,000 characters. A 10,000x compression ratio. Extrapolate that example out and there is no theoretical maximum to the compression ratio.
But that's when the data set is known and small. As the complexity grows, it does seem logical that a maximum limit would be introduced.
So, it might be possible to achieve 200x compression, but only if the complexity of the data set is below some threshold I'm not smart enough to calculate.
You also have to keep in mind that, the more you compress something, the more processing power you're going to need.
Whatever compression algorithm that is proposed will also need to be able to handle the data in real-time and at low-power.
But you are correct that compression beyond 200x is absolutely achievable.
A more visual example of compression could be something like one of the Stable Diffusion AI/ML models. The model may only be a few Gigabytes, but you could generate an insane amount of images that go well beyond that initial model size. And as long as someone else is using the same model/input/seed they can also generate the exact same image as someone else. So instead of having to transmit the entire 4k image itself, you just have to tell them the prompt, along with a few variables (the seed, the CFG Scale, the # of steps, etc) and they can generate the entire 4k image on their own machine that looks exactly the same as the one you generated on your machine.
So basically, for only
a few bitsabout a kilobyte, you can get 20+MB worth of data transmitted in this way. The drawback is that you need a powerful computer and a lot of energy to regenerate those images, which brings us back to the problem of making this data conveyed in real-time while using low-power.Edit:
Tap for some quick napkin math
For transmitting the information to generate that image, you would need about 1KB to allow for 1k characters in the prompt (if you really even need that),
then about 2 bytes for the height,
2 for the width,
8 bytes for the seed,
less than a byte for the CFG and the Steps (but we'll just round up to 2 bytes).
Then, you would want something better than just a parity bit for ensuring the message is transmitted correctly, so let's throw on a 32 or 64 byte hash at the end...
That still only puts us a little over 1KB (1078Bytes)... So for generating a 4k image (.PNG file) we get ~24MB worth of lossless decompression.
That's 24,000,000 Bytes which gives us roughly a compression of about 20,000x
But of course, that's still going to take time to decompress as well as a decent spike in power consumption for about 30-60+ seconds (depending on hardware) which is far from anything "real-time".
Of course you could also be generating 8k images instead of 4k images... I'm not really stressing this idea to it's full potential by any means.So in the end you get compression at a factor of more than 20,000x for using a method like this, but it won't be for low power or anywhere near "real-time".
Neurons work in analogue data, I'm not sure lossless algorithms are necessary.
Why should we? What's in it for us?
A job interview! (I wish I was joking).
The reward for developing this miraculous leap forward in technology? A job interview, according to Neuralink employee Bliss Chapman. There is no mention of monetary compensation on the web page.
Nothing, but then you could patent it and license it to anyone but elon :) are you motivated yet?
You can have a free "flamethrower" cigarette lighter. The company is bankrupt, and musk has a warehouse if the things he didn't sell.
I mean damn bro helping humans potentially walk again is a pretty big "for us" thing if you think about it in terms of humankind and not just yourself. Like imagine if someone were trying to cure cancer with the help of the public and you're all like "well what the fuck is in it for ME though?"
Imagine we all pooled our resources to fund medical research through taxes only for private companies to exploit the technology and jack up the prices…
A brain implant for rich people isn’t necessarily “for us”.
Oh but I'm not saying this out of selfishness, the problem for me is not the cancer cure in itself, but who is doing the research..
- the experiments on monkeys were questionable in method and nature, and led to death and madness;
- the other chip installed in a human has already lost the majority of connection wires;
and not to forget, it's not been specified how the public giving the ideas, would benefit from it. Musk is not exactly known as the phylanthropic kind.
That isn't at all their problem their problem is scar tissue buildup that they haven't even bothered addressing. Wtf are they doing talking about data compression when they can't even maintain connection.
You really think they only have one problem to solve? If that were the case this would be relatively easy.
There were rumors of that and a lot of other complications in the animal trials. I don't think we ever got proof, but a lot of irregularities that were explained away. Could be a lot more problems coming.
Cause there are always more patients... but more data will let them get more press when it enables more interesting demos.
Did they try middle out compression?
Already solved by evolution. This is the same problem as all of us have with visual data. We’ve evolved to need much less data transfer by doing some image processing first. Same deal. Stick some processors in there so you only need to transfer processed results, not raw data
Listen Elon, I have three words that will blow your mind: Middle out compression!
That's a lot more civil than the three words I have for him.
He's such a genius, why would he look for additional help? All these claims are such shit. Remember when Tesla would be fully self driving and we would all whizzing around in tunnels? Fuck this guy.
Tesla is a load of shit for sure but SpaceX and this Neuralink of it really does what its supposed to, actually contribute to humanity. Especially this.
Brain machine interface development has been around for a lot longer than nueralink. Musk is just better at getting his stuff into the headlines. Yes, the idea is good and beneficial to humanity, but then so are electric cars. That's part of Musk's grift. He latches onto something genuinely good and turns it into his pet project so that any criticism of how he does it can easily be deflected, because he's automatically the good guy just for being there at all.
I've got some of those bags you put your clothes in then seal with a vacuum cleaner, if that's any use
Surprised they haven't tried to train a neural network to find a compression algorithm specifically for their sort of data.
There's a ridiculous irony in the fact they haven't, and it's still ironic even if they have and have thrown the idea out as a failure. Or a dystopian nightmare.
But if it is the latter, they might help save time and effort by telling "the public" what avenues have already failed, or that they don't want purely AI-generated solutions. Someone's bound to try it otherwise.
They did, but then Elon insisted they add a virtual neuralink into it and now the neural network is braindead.
This seems more like a hardware issue than a compression algorithm issue, but I could be wrong
How do you send 200x as much data?
You don’t. The external system needs to run an approximation of the internal system, which the internal system will also run and only transmit differences.
There you go. Solved it. (By delegating to a new problem.)
They want to add compression to the implant?
And how does the brainwave data look? I'm sure they have some samples?
They want to add compression to the implant?
They're making their own silicon for their sensor so adding an on-die ASIC for a specific compression method sounds pretty attainable.
Just add 199 more transmistters
Have they tried Middle-Out compression?
Zombie signal not strong enough to make people think Elon is excentric genius and not an loud moron?