• 0 Posts
  • 34 Comments
Joined 1 year ago
cake
Cake day: July 9th, 2023

help-circle


  • If not vanilla Ubuntu, I’d still suggest trying an Ubuntu derivative like Linux Mint or POP! OS. Ubuntu has a huge community, so in the event you run into issues it’ll be easier to find fixes for it.

    What you’ll find is that Linux distros are roughly grouped by a “family” (my term for it anyway). Anyone can (theoretically, anyway) start from a given kernel and roll their own distro, but most distros are modified versions of a handful of base distros.

    The major families at the moment are

    • Debian: A classic all-rounder that prioritizes stability over all else. Ubuntu is descended from Debian.

    • Fedora: Another classic all-rounder. I haven’t used it in a decade, so I won’t say much about it here.

    • Arch: If Linux nerds were car people, Arch is for the hot rodders. You can tune and control pretty much any aspect of your system. … Not a good 1st distro if you want to just get something going.

    There are many others, but these are the major desktop-PC distro families at the moment.

    The importance of these families is that techniques that work in one (say) Debian-based distro will tend to work in other Debian-based distros… But not necessarily in distros from other families.




  • I DM for my kids. Our first dungeon ended with an ogre-mage sending undead after the party.

    The heroes made quick work of most of the zombies and skeletons - and critted the ogre-mage so fast that the dungeon was cleared in record time…

    And then the party’s necromancer (My oldest) decided to raise the ogre-mage. And made them his undead lich-thrall, with instructions to rule the dungeon until the necromancer’s return to the region…

    Huh.


  • It’s not as good, but running small LLMs locally can work. I’ve been messing around with ollama, which makes it drop dead simple to try out different models locally.

    You won’t be running any model as powerful as ChatGPT - but for quick “stack overflow replacement” style of questions I find it’s usually good enough.

    And before you write off the idea of local models completely, some recent studies indicate that our current models could be made orders of magnitude smaller for the same level of capability. Think Moore’s law but for shrinking the required connections within a model. I do believe we’ll be able to run GPT3.5-level models on consumer grade hardware in the very near future. (Of course, by then GPT-7 may be running the world but we live in hope).


  • I used to work for an imaging satellite company. And yes - spy satellites are crazy powerful. The real problem is one of bandwidth. Crazy powerful spy satellites are expensive - and there aren’t a lot of them.

    So everybody is competing for time on them. Satellite images have been traditionally expensive and rare. We web intelligence agencies have to take turns and sometimes miss important events due to scheduling or timing conflicts.

    The thing these new satellites offer is broad coverage. When you have a few hundred small-sats there’s just many, many more opportunities to have eyes on the part of the world you’re interested in.

    All that said, you want to pay attention to the resolution of the images. The place I worked for was providing imagery about 1-meter resolution. E.g. each pixel in the image corresponded to about 1sq-meter of earth. We figured this was a good compromise between image quality and privacy. Enough to count cars, see weather patterns, make out groups of people, but identifying any given person was right out.

    So if you see an imaging company throwing a bazillion imaging small-sats up - its worth checking what their reported resolution is. 0.5m means a real tall dude would still only be 2 pixels. But 1cm resolution means you could count their teeth.


  • Fwiw, I setup my pihole at home using docker. I run a full size desktop as my all-the-things server and use it as a docker host. Makes managing my services much easier.

    I could, of course, use an actual raspi for this, but I run a bunch of other services - including my plex host and file server - on the same machine. Using docker makes it dead easy to update my various services as needed and no worries about dependency Hell between them.


  • swordsmanluke@programming.devtoPrivacy@lemmy.ml*deleted by creator*
    link
    fedilink
    arrow-up
    10
    arrow-down
    1
    ·
    9 months ago

    pushes glasses up nose Ackchually…

    The recent CPRA regulation in CA has essentially mandated automated data deletion requests. Technically it only applies to CA residents, but it’s so hard to disprove residency that most companies will process requests from anybody.

    It only went into effect last year, but yeah - everybody I’m aware of has implemented an api for processing requests.

    I think $9/mo is pretty fair to cover paying for the engineering and infrastructure to support their ongoing integration efforts.

    That said, you could absolutely build something yourself that sends automated requests to every data broker you can find, but… Mozilla already knows where they are and will be looking for more. It’s going to become a game of whack a mole as companies that haven’t received deletion requests will have more complete (and thus more valuable) data sets.

    If you don’t want to just leave it on though - just this a couple times a year as a sort of spring-cleaning event should cut down your presence on ad rolls significantly.


  • I work in an advertising-adjacent industry. My company doesn’t collect data ourselves, but we do purchase and use advertising data on behalf of our direct customers.

    First off, there’s no single “advertising id” in use across the industry. Some companies make up their own, some companies don’t have one at all. Several companies just link by your email address.

    You may be interested to know that the CPRA legislation in CA from 2023 has made it a legal requirement to allow customers to request that businesses:

    a) disclose what data they have about you

    b) allow you to delete your data

    … and a few other things.

    Technically, this only applies to CA residents, but (dis)proving residency is hard enough that most companies will just accept your request regardless of where you live.

    If you poke around, you should be able to find a way to submit CPRA requests to any given advertising company to request to see your data.

    This comes with a big caveat though - the Stalker Problem. What if some asshole goes to AdSense and says “My name is totally Jane Doe, what do you know about me? Recent addresses, especially.” … That gets into scary waters quick.

    The compromise many places have landed on is to confirm what they know about a person, but not volunteer any extra info. E.g. “I’m Jane Doe - what do you know about me?” -> “We know about Jane Doe.” or “We know nothing about Jane Doe.” (and if you provide email addresses etc, those may be individually confirmed or denied.)

    There’s a new framework of intermediaries popping up that will automatically submit your info for deletion across the industry, so if you sign up for one of those you can have your data regularly cleared.





  • One thing I’ve been experimenting with is creating archetypal characters ahead of time and then handing each player a couple sheets to choose from. If they don’t like either one, they can grab a sheet someone else discards.

    I leave backstory details blank, but the basic mechanics of each character is already laid out.

    In theory, it seems like I’m setting myself up for fights with my players, but in practice it hasn’t been a problem.

    …of course this might be because my players are all either adults or preteens.





  • …yes, but that also has the trade-off of moving your rolls from a flat distribution where every value between 1-20 have equal weight, to a bell curve that peaks at 10.5.

    Many of your rolls are gonna end up right around that 10-11 mark as a result. Which can be fine! #alldicearebeautiful

    But it’s not gonna be a great drop in replacement for D&D. D&D’s skill checks are built around beating numbers that you’re not going to reach as easily with 3d6 vs a flat d20.

    Basically, more dice = more predictability and fewer wild swings of fortune. That is a more accurate model of reality… But arguably less fun in a game.

    Imagine the difference in dramatic tension in a game where the boss has 50 HP. In one scenario, you deal a consistent 5.5 HP each round. In the other, you deal 1d10 damage each round.

    In the long run, you’ll deal the same amount of damage in either system. But the randomness of a 1d10 creates more dramatic tension and excitement! When you roll a 1, it’s a crushing setback. A 10? Instant jubilation.