[HN Gopher] Learning happens in the brains of sleeping babes
___________________________________________________________________
Learning happens in the brains of sleeping babes
Author : Hooke
Score : 36 points
Date : 2022-01-26 14:10 UTC (1 days ago)
(HTM) web link (knowablemagazine.org)
(TXT) w3m dump (knowablemagazine.org)
| Buttons840 wrote:
| As a Reinforcement Learning enthusiast, I can't help but see
| humans learning as they sleep as a "batch". In RL you often delay
| updating the agent and only update it periodically, just like
| brains seem to do with sleep.
| cultofmetatron wrote:
| makes a of of sense too when you spend a of of time learning
| new things. I've. put time into video games or learning a new
| programming language and found myself initially struggling. Yet
| the next day, things I struggled with just make sense without
| much effort. I've gotten to a point where I view the time
| reading the material as a passive endeavor, I practice enough
| to feel like I'm not too stressed out but getting a bit of
| challenge. The real test comes the next day after sleeping on
| it.
| i_like_apis wrote:
| I think if an advanced adversarial ET ever came into contact with
| humans they would view this whole "sleep" thing as a huge
| weakness. Having a period of rest for maintenance, repair, and
| integration is understandable for an organism -- but why spend a
| full 3rd of available time on it?
| louthy wrote:
| Presumably because it requires a ton of energy, so shutting
| every other system down is the most effective way? (well, at
| least when our next energy fix wasn't so easy to acquire)
| i_like_apis wrote:
| I think the reason we sleep so long has a bit to do with our
| lack of predation. It might be more of an evolutionary
| accident than a necessity.
|
| Cats are another Apex predator who sleep a lot, but the fact
| that they do it in naps is probably an evolutionary remnant
| of when they were not so apex. Primates have been apex for
| longer (?) and have longer been adapted to sleeping in trees
| away from predators.
|
| I think you're on to something about energy. I suppose the
| longer an organism has the luxury of being able to spend in
| an energy-conservation mode, the more it will do so when
| evolved in an energy-intensive world. And
| maintenance/repair/consolidate modes will prefer to converge
| with these modes as that is a time when it is most safe to
| perform. So a long time spent sleeping is perhaps emergent in
| all apex predators who have energy-expensive biologies.
|
| But the substrate / environment of life matters too. Aquatic
| creatures have shorter sleep cycles because the dangerous
| environment requires being alert at all times. Apex predators
| resort to strategies like sharks, who switch half their brain
| asleep while the other is awake, or whales, who sleep for
| only up to 30 minutes to maintain breath.
|
| I think a silicone AI or other advanced life-form may evolve
| to spend similar amounts of time in maintenance mode
| depending on the energy economy of their environment, but
| they would do it at shorter timescales that would make
| competing with it impossible. And the level at which they
| could function in "sleep" mode would probably rival our awake
| mode by orders of magnitude.
| ricardo_navarro wrote:
| Interesting study. I wonder if they can find any links with how
| well babies sleep to the impact it has in adults' sleeping
| behaviors...whether we carry things from the patterns learned
| early in life?
| in_cahoots wrote:
| It's not clear to me, how can the authors be sure that babies
| actually learn in their sleep, as opposed to just performing
| better because they're not tired?
| PeterWhittaker wrote:
| Possibly at least in part because the EEG readings they
| describe are consistent with memory consolidation.
|
| I _would_ like to see more on this, though, on the whole
| question of the effect of rest.
___________________________________________________________________
(page generated 2022-01-27 23:01 UTC)