• 8 Posts
  • 438 Comments
Joined 1 year ago
cake
Cake day: July 21st, 2023

help-circle






  • It is kind of interesting how open machine learning already is without much explicit advocacy for it.

    It’s the only field I can think of where the open version is just a few months behind SOTA in all of IT.

    Open training pipelines and open data are the only aspects that could still use improvements in ML, but there are plenty of projects that are near-SOTA and fully open.

    ML is extremely open compared to consumer mobile or desktop apps that are always ~10 years behind SOTA



  • 2 axis solar trackers are much more efficient, but fixed installation beats them in cost/W in many cases.

    Any solar installation gets dirty, the question is do you save labor/equipment cost by having them cleaned by a single solar cleaning train, vs. tons of workers or automated brushes cleaning a large open field installation. Do you need to do cleaning passes after every train? Daily? Monthly? Yearly? Is there an intersection of efficiency loss and cleaning investment that is profitable?

    If you could install and maintain them in a fully automated way with just a few specialized trains, I can see why it might be an attractive idea. Question is how automated can you make it really? Do you need to fasten the panels down? How do you tie them into the grid?

    If the savings on installation, maintenance and cleaning offsets the loss in revenue from the suboptimal placement and dirt, it might work.

    I could see this working out if deployed on large scales, where the up front investment of developing all the specialized process and equipment, like trains, becomes a small part of the cost.

    Any such proof of concept installation of an unproven technology will be more expensive than if you really deploy it at scale.

    If rail didn’t exist today and we had to develop the first train and track and all the necessary infrastructure around it, the first 10km would be ludicrously expensive and would never pay itself off compared to the existing road network or shipping routes.

    It’s a finetuning and risk taking problem. Does the idea make sense in a vaccum? And does the idea work in competition with existing solutions? Is anyone willing to invest enough money to make it competitve?

    I hate it when extremely complex multi-variate problems always get judged based on one or two possibly negligable variables because of ignorance or intellectual laziness. Sometimes you can successfuly jugde things this way, yes, but rarely are things that simple.