you are viewing a single comment's thread
view the rest of the comments
[–] 234 points 1 week ago (34 children)

Do you have a few minutes to talk about our Lord and Saviour, Linus Torvalds?

  • source
  • hideshow 34 child comments
  • [–] 132 points 1 week ago (3 children)

    Just opened the thread to make sure someone was spreading the good word.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 20 points 1 week ago (4 children)

    Our Linus, who art from Finland, Hallowed be thy kernel.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 3 points 1 week ago (3 children)

    He's been in California for decades.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 5 points 1 week ago (17 children)

    If I don't find an anti-AI kernel fork I might be moving to NetBSD soon. That sucks, I hate moving to a new OS.

  • source
  • parent
  • hideshow 17 child comments
  • [–] 34 points 1 week ago (16 children)

    Linus has always prioritized results over ideological purity. Which is why Linux runs the world whereas BSD is still just a niche project.

  • source
  • parent
  • hideshow 16 child comments
  • [–] 11 points 1 week ago (1 child)

    I’m not totally convinced that his behavior in the past is a reflection on what is to come. He is at this point still refusing to answer questions about licensing LLM generated code and such and has apparently been obfuscating LLM commits. That doesn’t sound like somebody who is committed to results so much as it sounds like somebody giving into AI brainrot.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 1 point 1 week ago

    He's also human, for example he hypothetically could use AI to offset a disability like dementia - a totally reasonable emotional response, but also not who I want controlling global cyber security.

    Last I heard though all AI commits were a separate branch, I wasn't aware of any sneakiness.

  • source
  • parent
  • [–] 7 points 1 week ago* (9 children)

    Clearly he doesn't care about results because he slop coded with antigravity rather than reading documentation for a python side-project. I imagine his younger self would be cussing his older self out for shipping things he literally doesn't understand.

    The reason for his actions are much simpler: he got paid. He gets paid from the linux foundation which receives a lump sum from AI companies. In 2024 Linus said AI was just Hype, but then the foundation received millions in funds from the AI companies.

    If he cared about results he would have written the damn things himself.

  • source
  • parent
  • hideshow 9 child comments
  • [–] 3 points 1 week ago (8 children)

    I'm no vibecoder but between 2024 and now, LLMs also got much better. It might not have been the money that changed his mind.

  • source
  • parent
  • hideshow 8 child comments
  • [–] -1 points 1 week ago (7 children)

    LLMs either hit a wall or got worse, so no.

  • source
  • parent
  • hideshow 7 child comments
  • [–] 2 points 1 week ago (6 children)

    As much as I'd like that to be true, this is unfortunately not the case...

  • source
  • parent
  • hideshow 6 child comments
  • [–] -1 points 1 week ago (5 children)

    You're going to argue that some new never before seen trove of uncontaminated training data was discovered? Perhaps you think the AI Scaling Laws don't exist? Businesses that adopted AI dropped it or went out of business not long after, it's as useless today as it was in the 1980s.

  • source
  • parent
  • hideshow 5 child comments
  • [–] 2 points 1 week ago (4 children)

    No, I'm arguing that the models got better. I unfortunately have to work with them...

  • source
  • parent
  • hideshow 4 child comments
  • [–] -2 points 1 week ago (3 children)

    Pure fantasy, mate. The only thing these machines have proven effective at is psychosis.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 3 points 1 week ago (2 children)

    In what fantasy world are you living, that you don't believe that??

    I'm living in the world of software engineering and this is a fucking reality now... I don't like it. I really don't like it. I really would like to agree with you, but it's not the case, since they have introduced reinforcement learning a while ago, the models got quite a bit better, this has nothing to do with AI psychosis or anything.

    They are not superior for everything (fortunately, otherwise I would have to fear about my job), and I say that as a quite experienced software engineer, but for rather dull tasks they are certainly quite a little bit faster than manually programming per hand, although it's certainly more fun to program by hand even dull tasks...

  • source
  • parent
  • hideshow 2 child comments
  • [–] 6 points 1 week ago

    I'm not sure I'd call permissive licensing "ideological purity" compared to copyleft.

    Frankly, I think BSD loses because corporations are legally allowed to take and never give back.

  • source
  • parent
  • [–] 5 points 1 week ago (4 children)

    Didn't he jump on the AI bandwagon?

  • source
  • parent
  • hideshow 4 child comments
  • [–] 2 points 1 week ago (3 children)

    No, he said AI may be used for assistance in kernel commits, but the same code quality and comprehensibility as before AI is required.

    He cares about the output, not it you got some code snippet from ai or stackoverflow.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 2 points 1 week ago (1 child)

    And AI code is intrinsically bad because is fundamentally impossible to understand (yeah, I mean literally). You just can understand what you actually put your effort to comprehend, reason and write by yourself. They're just delegating their reasoning to a machine.

    Also, Why not LLMs?

  • source
  • parent
  • hideshow 1 child comment
  • [–] 1 point 1 week ago

    AI is just stolen code. It's often similar quality to human output

    I once tried to remake one of my projects using AI. When I gave it the requirements, it damn near recreated the project exactly how I made it. It was probably referencing my exact project, which would have been its training data, but it was a bit weird how well it did it

    It was not, "fundamentally impossible to understand" either figuratively or literally

  • source
  • parent