Lossless compression and decompression use identical modeling (prediction) steps and only differ in the coding step. It is the modeling step that is of interest to AGI. The only reason to decompress in testing the model is to verify that the coder doesn't have any bugs and is giving you the right numbers.
I think what you are describing is lossy compression. You find predictable patterns in the data and subtract them, leaving behind noise. The predictable part can be compressed losslessly. The residual cannot be compressed, so you either store it (lossless) or discard it (lossy). What you are testing now is your model of the human perceptual system. If you got it right, then you get both a good compression ratio and a decompressed result that looks and sounds like the original. Both are of interest to AGI. Lossless text compression is easy to evaluate. Lossy image, audio, and video evaluation requires humans to give subjective ratings. On Sun, Mar 7, 2021 at 1:22 AM Alan Grimes via AGI <[email protected]> wrote: > > BLARGH!! > > U know Lord of the Rings, Return of the King where the Eye of Sauron is > looking over the wastes of Mordor for Frodo? That's what the HOA is > doing to me over a bunch of outrageously petty subjective crap and is > now pulling horseshit like sending me threatening letters using the most > expensive services the us postal service offers and then billing me $50 > for those services on top of all other fines and penalties. I'm in the > market for a lawyer. I'm really sterssed out rn so I'm going to take my > mind off of it by taking out my frustration on all the superficial wheel > spinning I'm seeing about compression on the list these days. > > Lets look at some pictures, my appologies if you can't properly perceive > these as they are a test of common perceptual deficiencies. > > https://www.aoa.org/healthy-eyes/eye-and-vision-conditions/color-vision-deficiency?sso=y > > Ok... What is a perception anyway? > > A perception of a scene, or the equivalent in any other modality, can be > expressed as follows > > percept = SIGMA( F(), F(), F(), F(), ......., -<NOISE>-) > > Where F() is a function which yields a pattern which, when subtracted > from the signal, results in a simpler, less structured signal that does > not contain F() or any echo thereof, ie a compressed signal, or less > information-dense signal. > > Take the figure in the second row, second column. > > Here are some functions which compress that figure: > > F() = "3" > F() = Circle > F() = dot > F() = orange > F() = green > [and quite a few others...] > > In computing terms, we are de-serializing the perception and turning it > into a structured heirarchy of concepts. The brain actually renders, in > real time, everything you see! The human eye is pretty crappy as a > camera in all respects, but the brain only uses it to sample the > environment and construct the high resolution world model that we, > incorrectly, think is coming from our eyes! They eyes are only the > source of a feedback signal!!!!! > > So we are >>>>>> NOT <<<<<<< looking for a compression algorithm!!! We > are looking for a DECOMPRESSION algorithm but one where we feed it > garbage and then use a feedback amplifier to error corret the output > garbage to discover what the input compressed data should be! =P > > If you don't understand what I just said, you aren't trying nearly hard > enough, go away... If you can, however, then you know what you need to do. > > BLARG!!!! > > -- > The vaccine is a LIE. > #EggCrisis > White is the new Kulak. > Powers are not rights. > -- -- Matt Mahoney, [email protected] ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T5bda8b8be25887f8-M9d4d525a2f1aa27865e35c69 Delivery options: https://agi.topicbox.com/groups/agi/subscription
