> The process is important because what's getting washed away during sleep are waste proteins that are toxic to brain cells, Nedergaard says. Evolution found a nice little optimum in sleep that works pretty well- use less energy during a time of the day that isn't a very efficient time to be awake while also cleaning out the brain. This could be interesting for future research. Could we find chemical means to remove…
There's an interesting concept. I wonder what the trade-off would be. How much efficiency would it cost to clean the brain continually rather than in a dedicated sleep cycle? Would it even be possible to process these toxins while maintaining consciousness? I'm sure there could be huge benefits for many modern lifestyles to trade heavily on reaction speed and/or lightly on IQ for 50% more waking hours. Or even 25% mo…
So I imagine the trade-off isn't as severe as someone might guess; it's definitely theoretically possible to run entirely without sleep with the right hardware. For example, lets say sleep was fundamental to cognition. If you were willing to have redundant hardware, you could run two brains and just take turns sleeping them [1]. So that probably gives something of an upper-bound to efficiency.
The problem is that a lot of stuff is getting built on top of these things. Scheduled downtime is getting used by other adaptions. Things like handling 'garbage collection' or 'waste collection' inline get shunted away to the downtime to improve response times. This keeps happening. Over short periods there are going to be plenty of cases where being paralyzed isn't optimal. What if something shows in the night? Long term though if the creature is continually staying active when its at disadvantage that's really bad. So additional penalties accrue to force the issue.
Then the original 'purpose/benefit' starts getting lost. Things get messed up. Enough adaptions end up taking advantage of the fact that the downtime existed that now there is so much benefit to it for reasons unrelated to its initial creation that it starts becoming worth doing even at times where light levels aren't an issue.
So now when you try to mess with sleep, you're not just messing with one little thing. You're messing with a whole bunch of different entangled processes. One to reduce activity during periods where the sensors suck. Another to prevent continued activity when the sensors suck. Another that handles waste. Another that starts up ML algorithms during the downtime to take advantage of the freed-up processing. Another that does data backup. The muscles use the downtime to start construction. Basically all over the body everyone is starting to get to work.
It's like that tradeoff problem that gwern talks about with using notropics, where each one improves one thing but causes problems in other areas. So basically, it isn't a question of whether or not you need sleep from a theoretical process perspective. You don't, if you had a different design. But now sleep isn't just the reduce activity time because its night adaption. It's all tied up in everything else.
You shorten the amount you sleep? You die younger, because you cancelled the bodies maintenance period.
[1]: Which dolphins are kind of doing.