I've been working on a P50 since 2016; It is holding up.
Flaws:
- There is cosmetic degradation
- The laptop is gradually losing hassle-free hardware support from the latest popular Linux distributions. (Trackpoint crashes randomly in Ubuntu; Sleep/power management has never worked properly).
- I had to replace the keyboard after 3 years (the keys mechanically wore out and became loose). I had difficulty finding new replacement keyboards recently (although this is more an indictment of the UK's trade situation).
- The buttons on the main touchpad are slowly failing; Near as I can tell it isn't possible to replace these without replacing the entire "top case". I use an external mouse to slow the decline.
(Blitting unicode glyphs to screen is not hard or slow; In this project, the glyph rendering code turned into spaghetti to save on space. Cleaner rendering code is possible if using uncompressed bitmap fonts. If there is a concern about size on disk, you could only install the codepages needed for a given locale? But then again, unless you include the various pages of Chinese characters, we're still talking no more than a few hundred K to store all of Unicode as a bitmap font. True, handling combining modifiers probably would increase binary size and maintenance costs, but again... if it fits on an Arduino it's probably OK for the desktop. And if you're already OK with not supporting unicode, you're probably OK with supporting most of unicode except combining modifiers.)
If I understood correctly, I think I agree with your loose analogy. I've tried combining some error correction with ongoing re-learning. It stabilizes things against drift, but not indefinitely. This is all in an extremely preliminary cartoon sketch of a model. https://www.biorxiv.org/content/10.1101/2021.03.08.433413v1....
My understanding was that these experiments didn't completely rule out the possibility of a stable sub-population. Also, the olfactory bulb projects to other areas besides piriform, and those might support long-term stability. My extremely speculative unfounded conjecture is that piriform is actually a novelty encoder, and plays a role in encoding new memories in a changing environment. But, I think the data don't quite support this at this time.
I think part of the mystery is that computers can copy binary information losslessly, ad infinitum. So they can move information as often as needed with (almost) no error. We don't think neurons are as reliable, and excessive copying is expected to degrade a memory/representation. It seems like this might work for discrete categorical representations, whose activation is essentially binary, but would gradually fail for anything analog/continuous.
I think the point is that there are smells that we do remember for a lifetime, which must be explained. Either these stable memories aren't mediated by the cells in piriform that Schoonover and Fink recorded, or there is a mechanism that supports drifting yet stable distributed representations.
I'm (weakly) leaning in this direction as well. But, there are other proposed functions of dreams that are hard to reconcile with this. Some people think that dreams prevent overfitting and/or sample the negative gradient in a contrastive learning algorithm. Perhaps these are the same thing in end? But I'm not sure how it all fits together.
I'm working on this at the moment; Reconsolidation should work, but requires that representations/memories be periodically re-activated so that the neurons can re-learn how to encode them. Still very preliminary: https://www.biorxiv.org/content/10.1101/2021.03.08.433413v1....