EEVblog Electronics Community Forum
Products => Crowd Funded Projects => Topic started by: EEVblog on July 17, 2018, 11:38:20 pm
-
I'm sure they have done enough to con their investors (they obviously have seed investment already) into thinking this is actually order of magnitude game changing compression.
In reality it's bound to be mostly marking hype.
e.g. it's really really really hard to compress already compressed data like video, especially when it's lossless as they claim.
And of course it's on Indiegogo!
https://www.indiegogo.com/projects/delta-lambda-fit-1000-times-more-data-on-devices-apps-computers#/ (https://www.indiegogo.com/projects/delta-lambda-fit-1000-times-more-data-on-devices-apps-computers#/)
-
Wow, the more I look into this the dodgier it gets. I sense a video coming on...
-
Yep, they're claiming 1000X compression of 50GB blu-ray movies, = 1000 X total :bullshit:
I wouldn't be surprised if their names are fake, all the names I tried gave old film stars rather than any tech stuff.
-
They deserve a wack over the head with a statue of Shannon for a stunt like that. People still fund perpetual motion machines, so they won't stop funding things which which break the fundamentals of information theory any time soon.
-
Apparently real names, though they are going by their middle names for some reason.
Some analysis here, but coming up with the same question: "Who's the mark?"
https://www.metabunk.org/debunked-delta-lambda-compression.t9848/ (https://www.metabunk.org/debunked-delta-lambda-compression.t9848/)
-
They deserved to be compressed a thousand times.
"No, it's ok R2, you don't need to disable the garbage smashers on the detention level..." >:D
-
Well, if people are dumb enough to give them money...
To say I'm skeptical would be an understatement, however it doesn't actually break any laws of the universe that I'm aware of. The levels of compression currently achievable with MPEG on video are something that I wouldn't have believed possible 20 years ago. One must wonder how much better is actually possible to achieve.
-
Some analysis here
Which has a couple of boobies itself:
1GB in their claimed 0.033 seconds is a bandwidth of 30 GB per second. The world's fastest SSD drive tops out at 6.8GB/s. So they are claiming a speed that is five times faster than you can possibly even just read the data off the drive.
Who said anything about drives? Could be memory to memory, but suppose it does involve a drive - if the compressed part is on the drive and the decompressed in memory, the throughput needed would be 1000x less (so 3GB/s) and achievable.
Of course, it is a scam - he rightly points out the discrepancy in scaling - but if you're debunking something you do need to be right on all counts.
[Edit: No sooner said than: "1000x less (so 3GB/s)" D'oh. Should be 30MB/s]
-
I want to believe, but this is way too good to be true on so many levels.
-
They must have failed marketing this invention to NSA. ;)
-
Link seems to be removed.
Anyway this sounds like a flashback, end 99 a dutch "inventor" Jan Sloot claimed he could put tens of dvd's on a 64kB chip or a compression of 1:1000000.
The Philips second in command at the time Roel Pieper believed him and took his device to Philips Research where all the principal scientists claimed it was impossible and a fraud. They took it to the USA trying to get investment capital.
Jan never disclosed his algorithm and died prematurely just before the invention was sold. A book was written and blabla.
Latest I have heard it was not a compression algorithm but a code. My interpretation was that it was what DirectX is now for games, you have a large library containing the information in building blocks and need less code to draw what you want to draw. But hey since it was never ever revealed or reverse engineered it could have well been fraud.
https://en.wikipedia.org/wiki/Jan_Sloot
-
"Link seems to be removed." - All 3 links on this page work for me.
"the Sloot Digital Coding System, which could compress a complete movie down to 8 kilobytes of data — this is orders of magnitude greater compression than the best currently available technology in the 2010s."
"Despite the apparent impossibility of the encoding system, there were investors who saw potential. In early 1999, Dutch investor Marcel Boekhoorn joined the group. In March 1999, the system was demonstrated to Roel Pieper, former CTO and board member of Philips. Pieper resigned in Philips in May 1999 and joined Sloot's company as CEO, which was re-branded as The Fifth Force, Inc" https://en.wikipedia.org/wiki/Jan_Sloot
Why are CTOs and CEOs always clueless.
-
The name "Delta Lambda" makes me think that this could be some slant on incremental deduped backup algorithms (or in simple terms, rsync).
If you had a consistent large amount of data amongst all the users of the system, and were able to encode some lambda function which operated over that data set or portion thereof, to rip out bitstreams and potentially apply some delta to them to create a target bitstream which forms desired resulting data, then your "compressed" data consists of a series of lambda functions operating over a known data set and maybe any data that could not be otherwise described compressed by some standard algorithm.
I feel like there is something here that would make Kurt Gödel turn in his grave or something however.
-
Digital overunity machine! <3 <3 <3
Alexander.
-
To say I'm skeptical would be an understatement, however it doesn't actually break any laws of the universe that I'm aware of. The levels of compression currently achievable with MPEG on video are something that I wouldn't have believed possible 20 years ago. One must wonder how much better is actually possible to achieve.
Which MPEG specifically? As far as I'm aware, any sort of lossless H264 or H265 is not terribly efficient, you'd save about as much space zipping the video.
If you are talking about lossy H26x that is a different story and has nothing to do with OP.
Found semi related post on dropbox lossless: http://www.solveigmm.com/de/gewusst-wie/the-details-of-dropbox-h264-lossless-compression/ (http://www.solveigmm.com/de/gewusst-wie/the-details-of-dropbox-h264-lossless-compression/)
-
Well, if people are dumb enough to give them money...
To say I'm skeptical would be an understatement, however it doesn't actually break any laws of the universe that I'm aware of. The levels of compression currently achievable with MPEG on video are something that I wouldn't have believed possible 20 years ago. One must wonder how much better is actually possible to achieve.
It breaks the second law of thermodynamics. MPEG compresses well because it is lossy. It has been carefully designed to throw away what the human visual system is not very good at detecting, and that is a large part of a typical raw video signal. Lossless compression is only guaranteed to achieve a reduction in size if the data has some measure of predictability (e.g. is it known to contain a lot of repetition, and you can replace the repeating sections with one copy and a repeat count). Once you compress out everything which is predictable about the data, you are left with its raw entropy, and that can't be reduced further. For completely arbitrary data there is no predictability to begin with, and no compression is possible. This is basic information theory.
-
Wow, the more I look into this the dodgier it gets. I sense a video coming on...
Wow !! :popcorn:
-
Bollocks factor 9.99 Mr Sulu.
This shit appears every few years. The only time I saw anything which was any good was when someone showed me a 5 minute video with audio in a 1.44mb floppy. This was using wavelet encoding and the quality wasn’t great (think 90s realplayer shit) but it was far ahead of anything else out there. And it was genuine.
-
They resurrected a scam from the dial-up modem days?
-
I'm sure they have done enough to con their investors (they obviously have seed investment already) into thinking this is actually order of magnitude game changing compression.
In reality it's bound to be mostly marking hype.
e.g. it's really really really hard to compress already compressed data like video, especially when it's lossless as they claim.
And of course it's on Indiegogo!
https://www.indiegogo.com/projects/delta-lambda-fit-1000-times-more-data-on-devices-apps-computers#/ (https://www.indiegogo.com/projects/delta-lambda-fit-1000-times-more-data-on-devices-apps-computers#/)
I wonder if they are using middle-out compression.
-
It breaks the second law of thermodynamics. MPEG compresses well because it is lossy. It has been carefully designed to throw away what the human visual system is not very good at detecting, and that is a large part of a typical raw video signal. Lossless compression is only guaranteed to achieve a reduction in size if the data has some measure of predictability (e.g. is it known to contain a lot of repetition, and you can replace the repeating sections with one copy and a repeat count). Once you compress out everything which is predictable about the data, you are left with its raw entropy, and that can't be reduced further. For completely arbitrary data there is no predictability to begin with, and no compression is possible. This is basic information theory.
Ah I missed that this was claiming to be lossless.
Even though MPEG is lossy, I find it extremely impressive that it works anywhere near as well as it does.
-
It's conventional video which is an extremely inefficient format because it keeps sending (nearly) the same information over and over. Eliminate the repetition and you have a massive bandwidth saving, even with no compression.
Actually, this fact was exploited decades ago in the PAL colour decoder where the processing was split over two scan lines. Feasible with the tech available then because a simple 64us delay line was all that was needed. Modern codecs compare entire frames and only send the changes.
-
It's conventional video which is an extremely inefficient format because it keeps sending (nearly) the same information over and over. Eliminate the repetition and you have a massive bandwidth saving, even with no compression.
And that's what all the MPEG and other encoding formats do.
The problem with video is that (nearly) the "same information" varies enough for it to be a PITA due to noise on the image changing pixels, slight colour or lighting variation and reflections in the video image etc. If you have ever tried compress screen capture video vs supposedly "mostly still" video from a camera (even using tripod), you'll know that the screen capture data compress massively better than camera video does, because the frame-to-frame information is perfectly the same.
Still, these idiots claim 100x compression on already compressed video!
-
I can see an opportunity for AI to preprocess the video by removing noise while keeping wanted details. Getting it to work as intended isn't going to be easy...
The only time I saw anything which was any good was when someone showed me a 5 minute video with audio in a 1.44mb floppy. This was using wavelet encoding and the quality wasn’t great (think 90s realplayer shit) but it was far ahead of anything else out there. And it was genuine.
There are a few videos encoded in ASM that manage to fit several minutes in 64kB!
http://www.theproduct.de/ (http://www.theproduct.de/)
-
These scammers overhype their stuff so much. 1000x?
Make an algorithm which can losslessly reduce the size of any file by JUST ONE BYTE and you'll become a trillionaire.
-
There are a few videos encoded in ASM that manage to fit several minutes in 64kB!
http://www.theproduct.de/ (http://www.theproduct.de/)
That is because they're not videos. They're generating all of that on the fly using procedural generation. Instead of storing the finished textures or models, they'll store the steps needed to generate it from scratch using their custom graphics engine.
-
Make an algorithm which can losslessly reduce the size of any file by JUST ONE BYTE and you'll become a trillionaire.
Yep, then apply it recursively.....
Regards, Dan.
-
Reminds me of the good old days of Jules Gilbert on comp.compression, although the way he went away and came back in waves always made me suspect it was some kind of silly class project on how to troll usenet some professor used each year.
The Australians here might also remember Adam Clark (https://www.smh.com.au/business/compressed-into-nothingness-20100808-11qdh.html). I think his company later transformed into a shell company for a gold prospecting scam.
-
Wonder if they are using something like KGB Archiver (https://en.wikipedia.org/wiki/KGB_Archiver). This archiver can shrink 1GB to 1-10MB but man you are going to spend a good chunk of a day decompressing it
-
Was thinking about doing a video on this, but the campaign has gone nowhere, so I think best to just let it die.
-
Wonder if they are using something like KGB Archiver (https://en.wikipedia.org/wiki/KGB_Archiver (https://en.wikipedia.org/wiki/KGB_Archiver)). This archiver can shrink 1GB to 1-10MB but man you are going to spend a good chunk of a day decompressing it
No. You can't compress any random 1 GB file to, say, 10 MB. Some files will compress others will not. It just happens in a lot of cases that files we wish to compress (e.g. text) are easily compressed.
I don't think KGB Archiver was ever a big thing. It's reported as being very memory demanding at 256 MB, which probably means that no one has run this program in a decade.
If you are looking for a modern, powerful, memory hungry compression method then Bzip2 is a good candidate:
http://www.bzip.org/ (http://www.bzip.org/)
-
Wonder if they are using something like KGB Archiver (https://en.wikipedia.org/wiki/KGB_Archiver (https://en.wikipedia.org/wiki/KGB_Archiver)). This archiver can shrink 1GB to 1-10MB but man you are going to spend a good chunk of a day decompressing it
No. You can't compress any random 1 GB file to, say, 10 MB. Some files will compress others will not. It just happens in a lot of cases that files we wish to compress (e.g. text) are easily compressed.
The main problem these days is video, that's what need to be compressed, which is why they are making claims about video. But it bullshit, because the video is already heavily compressed, that pesky entropy thing. If it worked as they claimed it would be trivial to prove to everyone, and their startup would become an overnight unicorn (billion dollar company).
-
Wonder if they are using something like KGB Archiver (https://en.wikipedia.org/wiki/KGB_Archiver (https://en.wikipedia.org/wiki/KGB_Archiver)). This archiver can shrink 1GB to 1-10MB but man you are going to spend a good chunk of a day decompressing it
No. You can't compress any random 1 GB file to, say, 10 MB. Some files will compress others will not. It just happens in a lot of cases that files we wish to compress (e.g. text) are easily compressed.
I don't think KGB Archiver was ever a big thing. It's reported as being very memory demanding at 256 MB, which probably means that no one has run this program in a decade.
If you are looking for a modern, powerful, memory hungry compression method then Bzip2 is a good candidate:
http://www.bzip.org/ (http://www.bzip.org/)
bzip2 was great 10 years ago (and I used to sit next to Julian at Mozilla in 2009), but xz gets better compression and has been the compressor of choice for many projects since 2013 or so.
-
They resurrected a scam from the dial-up modem days?
I remember that! They deleted the file on disk and marked the sectors as bad. The "compressed file" was just a header and the FAT links to recreate the file.
-
That's hilarious. Wish I'd thought of that.