I wonder how consistent is the decompression and how much information is lost in the process.
i’d guess they could hyper optimise for “perceived difference” rather than data loss specifically… they do a pretty good job of generating something from nothing, so i’d say with enough data they’d probably generate a pretty reasonable facsimile of “standard” stuff
An LLM can't know what difference a person has perceived.
There have been a lot of studies done (and published) on what humans can and can't perceive. I wouldn't have much trouble believing that the LLM has access to them and can pattern match on the variables involved.
I wonder how consistent is the decompression and how much information is lost in the process.
i’d guess they could hyper optimise for “perceived difference” rather than data loss specifically… they do a pretty good job of generating something from nothing, so i’d say with enough data they’d probably generate a pretty reasonable facsimile of “standard” stuff
An LLM can't know what difference a person has perceived.
There have been a lot of studies done (and published) on what humans can and can't perceive. I wouldn't have much trouble believing that the LLM has access to them and can pattern match on the variables involved.
It's lossless: https://arxiv.org/pdf/2309.10668.pdf