5 ms·

It's not like computers are that good at multitasking either: - Lose our ability to focus for longer Well, multitasking is bad for the caches and any attempts

It's not like computers are that good at multitasking either: - Lose our ability to focus for longer Well, multitasking is bad for the caches and any attempts at speculation might have been ruined by the forced task switch - Prone to errors due to shorter attention spans Prone to errors due to Spectre and its followups. Some mitigations go as far as disabling hyper threading, due to leaks triggered by the really fast context switching between the two hyperthreads - Waste time during context switching Yep, CPUs need time for that too - Less productivity Oh yeah.. one old way of increasing throughput was longer time slices for less context switches. Eg the old Windows NTs differentiated on timeslice length in their server and workstation offerings. it's just that CPUs are still a helluvalot faster than us

kaba0 4y ago
People can only do single-threaded concurrency, we have absolutely no way at doing true parallelism on a cognitive level+ I really don’t see the parallel you are trying to draw, computers do what they are instructed to do with very very very little chance of a hardware failure. Not wanted operations can happen, but they ultimately boil down to human error. + We are otherwise a heavily distributed system which has many autonomous systems interoperating at the same time
- ThalesX 4y ago
  > Not wanted operations can happen, but they ultimately boil down to human error. This made me think. Is a cosmic ray human error because we were unable to prevent it damaging the system?
  - kaba0 4y ago
    How many cosmic ray bitflips happen? Even without ECC ram I’m fairly sure I have never seen any affect me, so there is that (even data hoarders with terabytes of data stored with e.g. ZFS seldom report problems due to that - when they start to see checksum errors it likely means a soon to fail harddrive instead) Also, if I’m not mistaken, in airplanes which are much more likely to get hit by cosmic rays at high altitude, they often use redundant CPUs that work in lockstep with each other. So.. in the end it is human error I guess? Handling unexpected cases should be done well as well.
- jgtrosh 4y ago
  The parallel they are drawing is for multitasking on single-core computers, for which performance is impacted somewhat similarly to humans. Of course, the same parallel cannot be drawn with multitasking on multiple cores; though other issues of communication between cores can (and have, extensively) be compared with parallelism between multiple humans.

kaba0 4y ago

People can only do single-threaded concurrency, we have absolutely no way at doing true parallelism on a cognitive level+ I really don’t see the parallel you are trying to draw, computers do what they are instructed to do with very very very little chance of a hardware failure. Not wanted operations can happen, but they ultimately boil down to human error. + We are otherwise a heavily distributed system which has many autonomous systems interoperating at the same time

ThalesX 4y ago
> Not wanted operations can happen, but they ultimately boil down to human error. This made me think. Is a cosmic ray human error because we were unable to prevent it damaging the system?
- kaba0 4y ago
  How many cosmic ray bitflips happen? Even without ECC ram I’m fairly sure I have never seen any affect me, so there is that (even data hoarders with terabytes of data stored with e.g. ZFS seldom report problems due to that - when they start to see checksum errors it likely means a soon to fail harddrive instead) Also, if I’m not mistaken, in airplanes which are much more likely to get hit by cosmic rays at high altitude, they often use redundant CPUs that work in lockstep with each other. So.. in the end it is human error I guess? Handling unexpected cases should be done well as well.
jgtrosh 4y ago
The parallel they are drawing is for multitasking on single-core computers, for which performance is impacted somewhat similarly to humans. Of course, the same parallel cannot be drawn with multitasking on multiple cores; though other issues of communication between cores can (and have, extensively) be compared with parallelism between multiple humans.