Infinite Compression Algorithm

Ok. I studied Wolfram's ideas. And he implies, that there is a simple cellular automata, that can produce any structure in existence. The only problem left is locating that structure in the automata's output.

Hey, FrozenVoid, can you do that?

If you know the entire length of the movie, you don't even need to consider "frames": only calculate the function of the R,G,B X-Y axises plots across the entire time of the movie.. 4kb for the desired goal functions, and 4kb for the "results" functions. I think the entire bytecode can be summarized with just a few bytes describing the entire formula compositions, like in chmod' rwx permissions are just primes 1,2&3.
I mean, what other math do you need than what most CISCs already have?

wow we just created actionscript and adobe flash

>>42
ActionScript is based of Frames: you can have infinite amount of frames per X time. E.g. in one second you could have 1 frame, or a trillion frames.
It also has some objects like circles, rectangles, lines, bezier curves, etc. It's too wordy, and not based on the goal: the monitors' RGB functions.

I could autistically expand it, but it's just understanding how to write formulas: passing math theory required here: you're effectively designing your general compression formula. None have beat LZMA yet.

>>41,43
so basically, you want to vectorize all the frames of the movie and come up with a function that interpolates between all the frames.

that's going to be much larger than the frames themselves, unless the movie is just 2 hours of a triangle rotating.

SVGs and flash already exist, how is this so revolutionary?

>>44
Vectorize the entire plot functions for plots R, G, and B across the X-Y which can simply be primes, across the entire length of the movie.
Similar like how in chess you have pawn, rook, knight, bishop, queen, and king, it's a 8x8 board, positions of legal moves can be described in functions of pieces and their effects to other pieces, and how they are affected by the opponent. The function/plot of let's say pawn white A2 to become queen, ignorant of black's moves, would look something like p(a2,a3,a4,a5,a6,[a7,b7],[b8,[a8,c8]]), which can be reduced to numbers pa2(1,2,3,5) since that pawn only has 4 possible choices in its entire goal to be queen, two of which are absolute if black never moved their pawns, rook, knight and bishop.
Why you'll need two formulas, since you'll have two functions contradicting each other, plotting the entire movie's on/off state of the RGB plot.
You are essentially drafting a "wack'a'mole" state formula: runtime of play is fixed, the "random"state of positions are plotted by sums of 1,2&3:RGB.

>>32
8kb is pretty tiny, but we don't count pre-shared data dictionaries/etc in compression. ie compressing the mnist dataset with a pre-trained neural net and not counting the net if it doesn't need to be transferred is acceptable.

So 8kb could probably do 1024 frames with 2^64 possible frame encodings?

Gives about a 30 sec clip at 30 fps, with room for ~18,000,000,000,000,000 different clips without sharing a frame

>>47
Something like a black and white movie would have RGB on definite values for their plots. Heck, I think the Z-axis would be 0/1 for on/off. Or a simple boolean stating the functions can be ignored in favor of reduced plot graphs.
>>48,49
You are still thinking in frames/time, think functions to state, why >>20 assumed Stool developed a generic quantum encoder.
I assume Stool had a fixed resolution, time was the variable, and RGB were 3seperate functions across the on/off state.

~1Mb per hour of video, but it might need a 1GB decoding model

>>46

Vectorize the entire plot functions for plots R, G, and B across the X-Y which can simply be primes, across the entire length of the movie.

I don't think you can come up with generalizable functions that compress any movie frame the same way a chess move can be compressed. The chess board comes with a strong prior--all the rules of chess. An image from a movie, by contrast, is completely arbitrary. There aren't any built in patterns you can exploit with the movie image.

What would be interesting is a superintelligence that could come up with a mathematical function that outputs an entire movie. For example, a video of a circle moving from the left to the right of the screen can be easily specified by the function for a circle with a varying horizontal coordinate. Imagine a superintelligence figuring out enough patterns on a larger scale over all the pixels in a movie such that it could reconstruct any frame by its index. I think that would be the closest you could get to infinite compression.

>>50
- http://jansloot.telcomsoft.nl/Sources-1/More/CaptainCosmos/Not_Compression.htm

So you'd get a supercomputer to compress every known video frame into a 1Gb (hopefully) model, and then you could store thousands of movies for 1Mb a pop..
But you'd still need to update the model yearly or whatever to get the latest movies

Sounds like using a URL minifier would be an infinite compression.

>>39

How large it is?

>>52
You can model all objects by ellipses, but that would still require a lot of motion data.

>>53
You will still need to define each possible frame somehow. And here we have that problem: frame index could be as large as the frame itself, if the set of all possible frames is too large. But yeah, in practice we have just 45000 hollywood movies, so it is possible to compress them all to just an index picking the movie.

If you stack frames of a movies, as layers of voxels, you can apply octree compression, because movies always have large areas of slowly changing gradients. In fact, modern video compression schemes, like AV1, indeed use octree compression:
https://hackernoon.com/av1-bitstream-analyzer-d25f1c27072b

The problem most of you are failing to overcome is that algorithms don't have to be perfect.

You probably use one every day that is sort of 'good enough', it's called Page Rank. The fact that one of the largest companies in the world's success is based on an algorithm that is 'good enough' still surprises me.

Maybe this is what Wolfram means, his new kind of science is something else, that we have yet to describe. A fabric we know is there, the empirical biological evidence is all around us; guess what; it's not perfect yet surprisingly effective.

>>59
Wolfram meant using computers to explore otherwise inaccessible math regions. But some math structures, like monster group, are hard to explore even using computers, because it is the size of our galaxy in atoms. Yet, at the same time, this monster group is intrinsically related with physics:
https://en.wikipedia.org/wiki/Monstrous_moonshine