sorted by: new top controversial old
Niederlande bauen souveränen Behördendesktop auf NixOS-Basis in c/de_edv@feddit.org
[–] 3 points 6 hours ago

Den Dönz werd ich mir wohl auch mal beibringen müssen, so gern ich auch alles auf Debian bastle aktuell.

Mein letzter Stand war, dass das Nix Projekt ziemlich AI- uns fash-friendly unterwegs ist. Hoffentlich kann ich mich über Lix oder Auxolotl etwas davon abschirmen.

  • permalink
  • fedilink
  • source
    •
    Firefox’s biggest priority: rounding corners in c/techtakes@awful.systems
    fun rule in c/onehundredninetysix@lemmy.blahaj.zone
    As a programming newcomer AND a dyscalculic, which ones are the best programming languages to learn that aren't tainted by AI and the best resources to learn them? in c/fuck_ai@lemmy.world
    [–] 5 points 9 hours ago* (last edited 9 hours ago)

    I can tell you the two languages I would learn right now if I had the spare cycles.

    1. Zig is a C-replacement language that took a no-AI stance early on. It doesn't have compiler-assured memory safety like Rust, but it does try to revamp structures (like alloc or error handling) in a way that makes it more obvious to the programmer what is happening. I'd love to learn more, but cycles...
    2. Gleam is a language that is part of the Erlang/BEAM ecosystem. It's a functional language, so it's a bit of a mindfuck if you come from a Python-/C-esque perspective, but if you like formal logic it might well be up your alley. It has a good web tutorial if you're interested. Neither Gleam nor its runtime Erlang-OTP have taken a stance on AI as far as I know (Elixir has a permissive stance), but at least I've seen no red flags as of yet. The people at Gleam just generally give me a vibe of being deliberate and careful with their code.
  • permalink
  • fedilink
  • source
    •
    Unionsländer drohen Berlin bei Linksregierung mit Konsequenzen bei Finanzierung in c/dach@feddit.org
    [–] 9 points 1 day ago (1 child)

    Ich weiß nicht was der betroffene Remmo beruflich macht, aber sich in die DMs von einem MdB zu schleimen der das Konzept Clan-kriminalität kritisiert und dann damit auf Insta anzugeben, wär schon so ein Arschlochmove den ich vom Mitglied eines kriminellen Clans erwarten würde.

  • permalink
  • fedilink
  • source
  • parent
  • context
    • On AI-generated articles in c/piefed_meta@piefed.social
    [–] 2 points 1 day ago* (1 child)

    Sorry for being late to the party, but I had to give my two cents on AI detectors. (specifically GPTZero & co., watermarking is different) I think it's a bad idea to recommend them. These companies are overpromising their accuracy and essentially selling false certainty. This will also not fade over time since the inaccuracy is baked into the tech.

    Currently the performance may be good for basic use cases, but recent papers seem to indicate you can still dazzle them by burning more tokens or explicitly prompting against detectors. But from what I understand about GenAI training* this is an eternal cat & mouse game, since the AI vendors don't want their output to be detectable.

    what I understand about GenAI trainingA basic technique of training a generative model is training a discriminative model alongside it. The gen model tries to synthesize output and the discriminative model tries to distinguish it from actual training data. The models are then trained alongside each other, building better and better discriminator models to sus out better and better generative ones.

    If these detector companies had some kind of breakthrough (better techniques, better datasets, whatever) to sus out generated output, they would be bought up to improve GenAI training. More likely, they don't. Maybe they can dial into some typical usecases for specific models and perform highly there, but that will fail to generalize for the edge cases.

  • permalink
  • fedilink
  • source
  • parent
  • context
    • sudo apt rule in c/onehundredninetysix@lemmy.blahaj.zone
    I really think that the term "Reverse Centaur" is underused in c/fuck_ai@lemmy.world
    [–] 0 points 4 days ago (1 child)

    I don't think it's impossible that the underlying tech of LLMs will be evolved into something undeniably useful. But I don't think "helping with Python" will be that.

    I think on !fuck_ai I can suggest, you could also just learn the relevant bits of Python? Learning to express your thoughts in code and understanding the tools you use feels way more empowering than relying on blackbox-extruded code.

  • permalink
  • fedilink
  • source
  • parent
  • context
    • I really think that the term "Reverse Centaur" is underused in c/fuck_ai@lemmy.world
    [–] 3 points 5 days ago (2 children)

    In my book socialism is about workers controlling the means of production, and historical socialists did not agree how that would work in detail.

    Libertarianism on the other hand is primarily about the rejection of personal restrictions, stuff like: "I should be allowed to freely copy and share media I have legally acquired." or "I shouldn't be forced to pay taxes." This individualist thinking makes it usually unable to adress structural inequality.

  • permalink
  • fedilink
  • source
  • parent
  • context
    • I really think that the term "Reverse Centaur" is underused in c/fuck_ai@lemmy.world
    [–] 3 points 5 days ago* (last edited 5 days ago) (16 children)

    But he also literally has a book titled "Information Doesn't Want to Be Free".

    That sure sounds like a zinger, but going off Wikipedia he does argue in that book for a loosening of copyright, which is pretty much what I meant. I don't hate those positions though I'll note they're more libertarian than socialist. He has a whole personal freedom + monopoly control shtick, which works great for me in a lot of scenarios, but doesn't quite work on GenAI tech.

    You can call me exceptionalist, but I think LLMs stand out because the adoption of the tech basically requires very capital-intensive and thus centralized operations for model training. This makes whoever runs them (likely capital or a state) a powerful arbiter of information space, maybe best comparable with algorithmic feeds in monopolistic social media.

    I think it is necessary to say: "Hey, we as a society shouldn't run things like that. There are better ways of distributing information that work just fine." A tech-neutral attitude can't achieve that.

  • permalink
  • fedilink
  • source
  • parent
  • context
    • I really think that the term "Reverse Centaur" is underused in c/fuck_ai@lemmy.world
    [–] 10 points 5 days ago* (21 children)

    Funny horse-butt-creature notwithstanding, Dotcorows AI takes have been rather mid. Quoting from Guardian...

    In this book, at least, he isn’t animated by the headline-grabbing concerns, whether existential risk or AI psychosis, deepfake porn or election disinformation, because those are unintended consequences. His target is the revenue model and the bubble it has created: “To be an effective AI critic, you need to strike at the source of AI’s power, which is the investment capital it attracts.”

    He isn't falling for AI doom and he is aware of the bubble, which puts him near the better end of the spectrum. But he's also unwilling to fully criticize community damage resulting from LLM use. That may be because he's fond of LLMs himself for his text editing and is repping the "it's just another tool" defense. So he isn't willing to go all the way when it comes to criticizing the global theft of labour that continues to power this industry. Information wants to be free, dontcha know?

  • permalink
  • fedilink
  • source
    •
    Nate Graham doubles down on sloppification in c/techtakes@awful.systems
    15
    What has the author of requests been up to? (tldr; talking to Claude) (awful.systems)
    submitted 3 weeks ago* (last edited 3 weeks ago) by to c/techtakes@awful.systems
    4 comments fedilink
     

    For reasons of my own to be revealed in due time, I was searching for the term "programmer mysticism" and found this essay by Kenneth Reitz: Programming as Spiritual Practice

    I was slightly intrigued at the start, since it was going in the "subjective experience of programming" angle I'm interested in.

    This understanding deepens when we recognize that consciousness itself might be a linguistic phenomenon. If consciousness emerges from patterns of language and mathematics, then programming isn't just creating tools—we're literally crafting the linguistic-mathematical structures from which conscious experience emerges.

    Well this concious approach to digital tooling sure didn't result in any scruples about LLM usage. Kenneth seems to have delegated most of his essay writing to Claude for some time (the article above is likely Claude-extruded). The piece continues to trudge through a basket of spiritual metaphors to hammer home how programmers should be more concious and do good things instead of bad things. This is all framed as a personal ethical issue without a shred of societal or structural analysis. Extra sneerable bit:

    The emerging understanding that LLMs contain humanity's digitized collective unconscious adds profound depth to conscious programming: when we collaborate with AI systems, we might be interfacing with the archetypal patterns that have guided human wisdom throughout history. This makes programming with AI a form of accessing collective spiritual wisdom rather than just advanced tooling.

    In another (likely slopped) article, linked in the same section, he goes into how he designed the API of the requests Python module to be far more convenient to use than the builtin HTTP resolver urllib2. The reasoning given strikingly reminded me of Intention Revealing Interfaces from Domain Driven Design. And I can personally confirm that library is a pleasure to use and a de-facto default for a reason.

    Took me a while to catch on to the sloppedness. I gave him some credits for requests and wanted to get a gist of the guy when I stumbled on his full-throated AI posts. Essays credited to Claude with himself as "editorial director". And of course this handy self-report: How I Write Now

    I guess this doesn't stand out that much by itself. It did kind of hit home when he describes in detail how his AI use is linked to his history of mental health problems, which he was quite public about. He describes being moved to tears by LLM writing, seeing "his thoughts" clearly articulated for the first time after years of brain fog. He made this accessibility argument explicitly. I won't waste my breath sneering about it. He gives plenty of fuel to that himself.

    10
    Build Golf (very rough post, low expectations only!) (awful.systems)
     

    Build Golf! Code Golf, but for Builds! Uhhh, that's sort of it.

    I remember being very slightly intrigued by the concept of code golf a long time ago. But I mostly took notice of people taking it in very esoteric directions, and even before that it was pretty much orthogonal to writing the kind of useful code I care about.

    But just now my brain rubbed 2 neurons together, and made the connnection to another problem that has bothered me quite often. How do you do a minimal build of some thing without some "helpful" 300 LoC boilerplate. I want like, 2 files with 3 lines each and one command to run.

    Okay, there will be a couple complications to consider to avoid this going down a hacky direction like Code Golf. Like, how much tooling do you depend upon? Do you have to install a package? How many (hundred) packages does that depend upon? (Looking at you React.) But the general idea is, how to you get (for a specific task or category of tool) in a minimalist way from plain-text in git to running machine code? (or like, HTML if you're building for the web) It's a challenge of taste more than discrete optimization! If tools will reign over others in this sphere, it should be because they cater to refined tastes!

    Soooo, that's my prompt. Ideas in all directions welcome. Concept for actually practically realizing this? Overly complex unworkable scoring rules? Personal fave minmal project templates? Feel free to evolve on this, or not, I'm not your boss.

    16
    Open Slopware Appreciation Post / Why do FOSS notes projects suck so bad? (awful.systems)
    submitted 2 months ago* (last edited 2 months ago) by to c/notawfultech@awful.systems
    22 comments fedilink
     

    Dear god. Why are note-taking projects such a magnet for slop bros?

    I'm on a bit of a quest to build up some self-hosted services for me and friends & family, partially because Ad- & AI-infested big tech slop was making me sick and also to become a better sysadmin. General requirement was notes available anywhere and decent editing on mobile.

    I was about to rip my hair out, when every session for weeks digging into options ended in frustration. Every project with somewhat decent features had some meh aspects and became worse the longer I looked. And there are so many of them. You can drown your self in "Best FOSS Note-taking App in $(date +%Y)" listicles. Here's a list of over a hundred projects.

    I suspect it's a bit of the shopping list problem. It's such a basic idea, that every ambitious clueless dev has it once while picking a hobby project. But note-taking also definitely has a place in toxic self-optimization / hustle culture. Some of the more developed project strike me as that.

    I won't mention every basic note-talking project with a CLAUDE.md that I stumbled on, but here's some of the more disappointing lowlights:

    • Joplin main developer was one-shot. Added agent instructions to the repo this year and some provisions for his personal vibecoding rig to .gitignore. Also it's not a bug if the AI chat button is visible even if AI features are disabled, because how would people find out about the AI features then?
    • Notesnook seems okay-ish, but they're part of the mall ninja club. Business partnership with Kagi & Tuta.
    • Standard Notes was the least suspect to me. Later found out they Copilot-Review all of their PRs. (same as Notesnook btw)
    • I even considered going with a self-hosted binary of Obsidian. They don't get instant-disqualified by promoting themselves as AI-anything. And if you don't have a project you can trust, you might as well throw yourself at the most established option and pray.

    I was running out of places to look for alternatives, when I remembered Open Slopware. You can guess how shocked I was, when note-software made up one of the biggest sublists in there. Through the shortlist of recommendations I ended up looking at Flatnotes. That project might be just enough notes for me.

    view more: next ›