> Through systematic quantification of distinguishable conscious states and their historical dependencies, we establish that the minimum information required to specify a conscious state exceeds the physical information capacity of the human brain by a significant factor.
Distinguishable conscious states doesn't mean every historical record needs to be present and fully encoded.
It can, and very likely is, lossy - in fact, it certainly is very lossy. So the "calculated" "bit-length requirements for representing consciously distinguishable sensory "stimulus frames" are bogus.
While "Distinguishable experiences require distinguishable physical states" is a pretty basic and obviously true statement, the idea that we carry all of our historical experiences, and do it faithfully, is bullshit.
> "If conscious states necessarily incorporate their historical antecedents, then the minimum information required to specify any given conscious state must include information about all preceding states."
No, it just needs to carry SOME information about SOME preceding states, potentially lossy and heavily compressed.
For a simple math example, a neural network can be fed terrabytes of data, but the end result is just some weights that encode "memories", "features", "functions" etc in the data.