• 0 Posts
  • 111 Comments
Joined 2 年前
cake
Cake day: 2023年7月10日

help-circle


  • I just upgraded from Win10 to Debian on my primary/gaming desktop; every game I’ve played in the last 4 years in my Steam library is listed as compatible, most things I do on my computer are either a website, a game, or an application with a native Linux version or good wine compatibility.

    As a personal user who does a wide variety of things on my computer, I don’t really need Windows anymore. I’m dual booting for the time being as I configure things, but I’m quickly running out of reasons to use the old Win10 partition. And all the weird slowdowns it was exhibiting? Totally absent in Linux. I expected performance to be a wash, but it’s noticeably improved.






  • This is one of many examples of a class of problem where the technology is the easy part. There’s room to improve the tech certainly, but the technology sufficient to solve the problem is already well understood.

    The hard part is how to get people to actually do the necessary changes. To consume less, get fewer gas cars on the road, increase the amount of nuclear, hydro, solar, geothermal, and wind in the grid, and minimize coal and gas use. To reduce land use by cows, and increase land use by trees and native plants.

    But maybe AI is the secret here. We have tools that are in the hype moment whose training data already contains several reasonable solutions to climate change. Maybe if AI “finds” the solution to climate change, people will finally listen



  • Certainly, some interesting developments have happened, and we’ve realized our old models/thinking about progress towards AGI needed improvement… and that’s real. I think there’s a serious conversation to be had about what AGI would be, and how we can know we’re approaching it, and when it has arrived.

    But anybody telling you it is close either has something to sell you, or has themselves bought it.




  • Critically, the people who build these machines don’t typically update drivers to port them to a new OS. You buy a piece of heavy equipment, investing tens, or maybe even a hundred thousand dollars, and there’s an OS it works on, maybe two if you’re lucky. The equipment hopefully works for at least 20 years, and basically no OS is going to maintain that kind of compatibility for that long. Linux might get the closest, but I’ll bet you’re compiling/patching your own kernels before 20 years is up.

    This kind of dynamic is unavoidable when equipment vendors sell equipment which has a long usable life (which is good), and don’t invest in software support (which is them being cheap, to an extent), and OSes change enough that these time horizons likely involve compatibility-breaking releases.



  • They could probably pass a law that companies within California pay taxes to the fed through an escrow account held by the state, pass the data through so if nothing crazy happens, it doesn’t really affect anybody, and get enough companies to comply that Trump would back down.

    The courts would eventually decide whether or not it was legal, but by then Trump will have moved off to another target to performatively attack while pretending to help people who aren’t multi-millionaires while siphoning their money instead.


  • It’s a structural challenge more than a fallacy, but I don’t entirely disagree. This sort of thing works best when one of two things is true, there’s some way for people to organize, or it’s relatively small and there are real options.

    The former clearly isn’t true here, but I think the latter is. There’s a lot of companies trying things with AI, and some are working better or worse. This particular use is relatively small, and I think the downside of doing it is also small in the short term. (This is a giant red flag, avoiding a red flag isn’t a large cost)





  • I suspect there’s a tendency of experts in something to think of people who do it narrowly as people doing at least as much as they are.

    The people who have a bunch of docker services, or complex multi-machine infrastructure are self-hosted software users, and probably in that 1-2% range. People who heard piholes are useful, so they bought a pi 3 and set it up are self-hosted software users. Somebody using an old desktop they got on Facebook marketplace for running Plex media are self-hosted software users… and so on. So are the people in their houses, some of their friends and family.

    Using that inclusive definition, being closer to 10% than 1% makes sense to me.