• 0 Posts
  • 12 Comments
Joined 2 years ago
cake
Cake day: June 12th, 2023

help-circle

  • Goodhart’s law strikes again.

    They can’t tune their process for ‘win election’, because that’s only one sample every four years, and it’s a binary value.

    So instead they tune it for ‘raise campaign funds’ as a proxy measure for ‘win election’, and that’s vastly more responsive; they can optimise the crap out of that.

    This also means that a bunch of influential people are able to skim significant amounts off the top, so they’re not minded to change it. They’re stinking rich so they don’t have to care about the actual political outcome - and the more people are suffering, the more they’ll donate.

    The trump win was a massive windfall for the next cycle of fundraising.








  • I know you’ve heard it, but seriously, the biggedst hurdle to learning programming is having a solution looking for a problem. You can learn all kinds of stuff in the abstract, but your brain will either handwave over it or bounce off, if it can’t contextualise the knowledge into a problem-solving tool.

    A source of small but genuine needs is the scarcest resource.

    I know the world has moved on to big cosy high-level languages that let you think semantically about everything but honestly I’m still a big fan of starting with C; it is something of an austere discipline, but it gives you a boots-on-the-ground view of the problems you’re working with, and I think that really helps shape your understanding of what you’re dealing with.

    (you could do a fair bit worse than the game ‘human resource machine’ for a slightly gentler introduction, then pick up some exercises in C, then move to python or similar - it will let you appreciate the cosy high-level stuff, and have a better feel for why you’re doing stuff)