Could someone please saw down this rifle barrel for me?
not only does it convert back losslessly it also has the noise removed
— moderate rock (@lookoutitsbbear) May 27, 2024
I am not an electrical or comp sci engineer but I have had some experience on the electrical side of things lately and have started becoming familiar with signal processing, so bear with me for minor mistakes if you are one of the nerds familiar with it. For those of you who are too 
or not deep enough on the 
spectrum to understand signal processing, all you need to know is this: noise is not data. It is never data. A good signals engineer must attempt to minimize noise as much as possible.
Neuralink put out some request recently asking for some insane level of compression on a file (wanting 200x compression) with a transmitted signal, and that this must be lossless (meaning the original file data can be perfectly recovered from the compressed file data). Neurodivergents online immediately set out working on this to determine what the max amount of compression is. Enter our king and hero, moderate rock (user @ lookoutitsbbear).
MR states that so far he's found a way to compress the initial data file by a factor of 4.1, the most seen so far.
Curious as to how, people start asking how well this actually works.
https://x.com/lookoutitsbbear/status/1794962035714785570
And this is where all heck breaks loose. See, Neuralink had technically asked for the original file to be compressed without any data lost. To the average midwit and/or software engineer, this means taking the original file and just making it smaller. But MR did what any good signal engineer would do, and worked on filtering the signal to get rid of unwanted and unnecessary information (the noise) so that he could do a better job of compressing the data. Midwits do not understand that noise is not relevant for a signal's information. And because of this, MR has sinned and for this, he must be dunked on. So the beatings commence.
"Heh kid, just google it. You idiot. You moron."
There are multiple midwits continuing to repeat the "noise is data" line, as though repeating it makes it true.
Once again, noise is not a relevant part of a signal.
The midwit bonanza continues, once again acting as though if noise is an important part of a signal.
Quote tweets even gain major traction dunking on him even though they're all wrong.
MR makes a quote tweet that at least gains traction with people who understand him, and the semantics argument becomes a bit more present.
Yes, on a technical level it is not lossless, as MR removed "information." But the "information" he removed is not relevant. The original signal present in the original file is recoverable after the compression. So he has done it correctly. MR also points out that the Neuralink engineers are being stupid because they should be working on getting rid of the noise before working on the compression.
Of course, the midwits can't admit they're wrong and continue to argue the semantics.
Someone who actually makes a decent enough analogy to understand what he's getting at.
There is a lot of continued insults and attempts at dunks here (too many for me to add at this point) and midwits continue try to dunk on him and he continues to shrug it off.
Adding to the hilarity, a PhD looks at the Neuralink thing and comes to the same conclusion on his own, that removing noise from the signal will help compression.
https://x.com/CJHandmer/status/1795486204185682315
https://x.com/kindgracekind/status/1795577979952845220
(some cope in the replies of this one that MR was "wrong" because of how he stated it, even though he wasn't)
Finally, our king decides to take a rest, having survived his beatings and coming out stronger 
So remember dramatards: your average midwit has no clue what they're talking about, software engineers should stick to learning to code, and you should always listen to the neurodivergent savants 
Jump in the discussion.
No email address required.
Biological noise is different from statistical noise. Biological noise is like I am comparing the expression of mRNA between two cell lines and in these cell lines some cells may be at different points in the cell cycle or slightly different environmental conditions (touching another cell, touching a surface, free floating ect) which we cannot expirementally control. For instance, at different points in mitosis a cell will have double the DNA and will be expressing different RNA compared to a cell at rest. We cannot reasonably control this in a living cell line without adding more uncertainty, and if we aren't interested in mitosis, you have to find ways to mitigate the noise and just get the signal. On a biological level this will never be perfect as biology has no reason to conform to our ideas about stats and any cutoff you do may always be removing some biological information at the margins (ofc you can do stuff like power series to exponentially expand the gulf between significant and insignificant data).
Anyway, this is just me saying that if the end goal of this is to be integrated into Neuralink this will inevitably become a biological problem rather than just a programming or engineering problem. As soon as biology gets involved, everything gets a lot more uncertain and annoying.
!codecels Dicussion on how the inclusion of biology makes the concept of noise and randomness more difficult
EDIT: I assume this is something where Neurallink is trying to compress neural signals since theyve reached a processing plateau and I am not a neurologist or bio psychologist I am a geneticist and microbiologist who does data analysis, so feel free to roast me if I'm off base
Jump in the discussion.
No email address required.
I'm not sure if you're cuter or smarter but you're definitely a smart cutie.
Jump in the discussion.
No email address required.
Aw thank you
Jump in the discussion.
No email address required.
You're welcome!
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
In biology noise is solved by sampling large numbers of cells per condition and averaging. You can also safely ignore anything irrelevant to the pathway you're interested in. For Neuralink a neuron firing/not firing is a very obvious and easy to detect state which makes things much simpler than the average gene expression study.
Jump in the discussion.
No email address required.
I used that example but say gene set enrichment analysis is an example of information being lost at the edges if significance. Its why there is now push back against thresholded cowtools vs those which use power scaling for pathway enrichment
Jump in the discussion.
No email address required.
How bout you enrich THIS
Jump in the discussion.
No email address required.
I wish i could never think about silence and just goon all day but sadly having a top of the line vr goon rig costs alot of money
Jump in the discussion.
No email address required.
More options
Context
More options
Context
I've never put much stock in enrichment analysis, a typical "data scientist" can usually find a way to get whatever result they want by fiddling with the cleaning parameters.
Jump in the discussion.
No email address required.
Enrichment analysis is fine if you use it as a jumping off point for lab work
Jump in the discussion.
No email address required.
Yeah, we both know nobody bothers to do that.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
Jump in the discussion.
No email address required.
More options
Context
Would the concepts of signal and noise in radio transmission be closer to biology or statistics based on how you've distinguished them?
Jump in the discussion.
No email address required.
More options
Context
The biggest fallacy is these nerds think a human being could be basically drag and dropped like a skyrim furry porn modlist.
Jump in the discussion.
No email address required.
More options
Context
More options
Context