[FRIAM] multitasking

[uǝlƃ ☤>$]

Right. The provenance of the use of "multitasking" in this thread is from the discussion of, and objection to, *tasks*. (Acknowledging Dave's masterful pirouette from arguing against [multi]objectives to arguing for [multi]tasking [⛧].) Unless the contexts for any 2 well-defined tasks can be unified, the context switch is pure overhead. And, usually, the contexts can be divided into parts, some of which are common to multiple tasks and some of which need to be switched out. I suppose that's the essence of different models of parallelism (e.g. SIMD) and multi-user operating systems. (Still waiting for a HURD or Plan 9. [sigh] Now I have to check up on new OSes. Qubes, anyone? <https://www.qubes-os.org/>)

Measures of difficulty in context switching may provide a marker for how *sharable* a task is (or can be). Sure, to some extent, it depends on the hardware (e.g. the particular human or the particular type of computer). But smeared out over, say, all standard humans with 10 fingers and 10 toes, and given 2 contexts of similar complexity, if context_1 generally takes longer to swap in/out than context_2, then we could argue that context_2 is "more human" than context_1. Similarly, some computation is more suited to the GPU than the CPU.

Narratives are inherently serial ... diachronic, more suited to the CPU, where context switching is fundamental, at least compared to more parallelizable things like POSETs. It wouldn't be surprising if people who believed humans were fundamentally narrative tended to disbelieve in human multi-tasking. It seems contradictory or paradoxical for someone who believes people are fundamentally story-tellers and, yet, also believe people are parallelist.

And *that* you can run on a treadmill at all says something about your architecture. I absolutely despise treadmills ... they violate everything I know (and hate/love) about running. What kind of monster are you?


[⛧] I suppose we could argue that objectives and tasks are different things. But my counter would be that they're something like [near]duals, or there's something like a Curry-Howard correspondence between them ... objectives are the initial and target state and tasks are the paths through the state space that connect the initial and target state, perhaps even some kind of path integral. An objective with no (possible) path toward that end is not a well-formed objective.

On 6/1/21 8:20 PM, Marcus Daniels wrote:
> I think the claim that people can't multitask is simply that there is distinction between short and long term memory.
> A computer can take an interrupt and retain and later restore its register file exactly.   This can happen thousands of times a second or years after a task suspension.  
> If it happens too frequently, progress will slow, but it only with very, very small probability will it fail (e.g. power fluctuation or cosmic ray strike).   But at least I need to take a quite a bit of time to restore the narrative to resume a detailed task.  Usually I avoid doing so until I know I will be free of distractions.   Things that I can easily restore are practiced or easy or meatspace things.
> 
> Incidentally, for running it depends if it is intervals or not.   Below anaerobic threshold, I can almost forget I am running.   On a treadmill, I can watch TV but I don't watch anything with a complex plot or with lengthy dialogue because I miss words or space out for periods.   Above it, I pay attention to the pain.  :-) 

-- 
☤>$ uǝlƃ