3 ms·

The elevators in Taipei 101 have to be among the most advanced, and yet it was insane how long I had to wait for one, how seemingly irrationally it went picking

The elevators in Taipei 101 have to be among the most advanced, and yet it was insane how long I had to wait for one, how seemingly irrationally it went picking people up, and what a mess it was figuring out your floor if you were using one of the double-decker elevators (my point is : the computer should figure it out for me). I’m terribly unimpressed by people programming real elevators.

fragmede 4y ago
Fun fact, elevator capacity is the limiting factor on skyscraper size. Taipei 101 (which has, creativity, 101 floors) has ultra high speed elevators, capable of running up to 37mph and it is the world's fastest elevator. It'll go from the 5th to the 89th floor in 37 seconds. The programming may seem haphazard, and it might actually be, I've never been there in person, but the elevator aren't bad for lack of trying. The new elevator system where you tell it the floor and it tells you which elevator to go to is part of the programming optimization.
- CobrastanJorji 4y ago
  I wonder if this is a scenario where it would make sense to advertise competence by offering more information to the user. Imagine a big visualization on the ground floor showing all 101 floors, the location of each elevator, which floors are waiting on people, where each elevator plans to stop, estimated ETAs, etc. It'd give waiting folks something pretty to idly study, and it'd also provide reassurance that the elevators are all doing quite a lot of work in a reasonably intelligent way. You'd probably need to balance this against some privacy concerns (maybe it's a bad idea to show people exactly when someone in the penthouse gets into the elevator).
  - Nextgrid 4y ago
    Short of privacy edge-cases, this would be a great idea. Dump the raw state of the elevator scheduling algorithm for the guests to look at while they wait.
  - fragmede 4y ago
    I think you're right. Part of the ML algorithm used to drive lifts is to guess and learn when elevators predictably need to be where (eg morning commute or lunch time), but ML is horrible opaque and everyone thinks they could do better.
- ouid 4y ago
  Isn't it also, for much the same reason, the limiting factor on "linear city" length? https://www.neom.com/en-us https://www.neom.com/en-us
  - samatman 4y ago
    For a linear city it would be practical to shunt cars off and onto the main line, so that throughput is a relatively constant function of safe stopping distance between cars in the event of emergency. So people traveling a short distance don't slow down ones traveling a long distance.

fragmede 4y ago

Fun fact, elevator capacity is the limiting factor on skyscraper size. Taipei 101 (which has, creativity, 101 floors) has ultra high speed elevators, capable of running up to 37mph and it is the world's fastest elevator. It'll go from the 5th to the 89th floor in 37 seconds. The programming may seem haphazard, and it might actually be, I've never been there in person, but the elevator aren't bad for lack of trying. The new elevator system where you tell it the floor and it tells you which elevator to go to is part of the programming optimization.

CobrastanJorji 4y ago
I wonder if this is a scenario where it would make sense to advertise competence by offering more information to the user. Imagine a big visualization on the ground floor showing all 101 floors, the location of each elevator, which floors are waiting on people, where each elevator plans to stop, estimated ETAs, etc. It'd give waiting folks something pretty to idly study, and it'd also provide reassurance that the elevators are all doing quite a lot of work in a reasonably intelligent way. You'd probably need to balance this against some privacy concerns (maybe it's a bad idea to show people exactly when someone in the penthouse gets into the elevator).
- Nextgrid 4y ago
  Short of privacy edge-cases, this would be a great idea. Dump the raw state of the elevator scheduling algorithm for the guests to look at while they wait.
- fragmede 4y ago
  I think you're right. Part of the ML algorithm used to drive lifts is to guess and learn when elevators predictably need to be where (eg morning commute or lunch time), but ML is horrible opaque and everyone thinks they could do better.
ouid 4y ago
Isn't it also, for much the same reason, the limiting factor on "linear city" length? https://www.neom.com/en-us https://www.neom.com/en-us
- samatman 4y ago
  For a linear city it would be practical to shunt cars off and onto the main line, so that throughput is a relatively constant function of safe stopping distance between cars in the event of emergency. So people traveling a short distance don't slow down ones traveling a long distance.