AIbros...

Ape Out Shirt $21.68

Tip Your Landlord Shirt $21.68

Ape Out Shirt $21.68

  1. 1 month ago
    Anonymous

    >pseudomathematical texts...a threat to the sound development of mathematics
    He'd know all about that wouldn't he

    • 1 month ago
      Anonymous

      Inter-Universal Teichmüller Theory is correct it does prove abc conjecture. But Black folk like you cannot comprehend it.

      • 1 month ago
        Anonymous

        Ok Mochi, stop posting and go watch some hentai you sick Japanese frick

  2. 1 month ago
    Anonymous

    AI is quite revealing of how conceited philosophical thinkers are, not just in these quotes but anyone with a philosophical inclination that I've spoken to echoes these same ideas, if you ask them to explain their rationale you quickly find that nothing is there, it's just ultimately cope, they don't even explain why they dislike AI so much
    I think that philosophy had one chance to be somewhat relevant with AI, in helping it advance, but most of them turn their backs on it and retreat to the fanfic reality bullshit that most philosophy lives in now

    • 1 month ago
      Anonymous

      >they don't even explain why they dislike AI so much
      It's modern geocentrism. Humans need to fell special one way or another. Our intelligence/creativity was one thing we had left to make us feel superior. If anything threatens that, it causes delusions and the brain's survival mechanisms try to "reason" their way out of it.

      • 1 month ago
        Anonymous

        1. we don't have anything close to real "AI"
        2. why would humans using their creativity to build AI somehow mean that humans are not creative? It's the opposite.
        3. your entire post is based on the reductionist view of "the brain as a machine/computer" which has been losing credibility for decades

  3. 1 month ago
    Anonymous

    I hate LLMs dogshit research so much. Please kill it soon.
    I want to go back to doing real AI instead of seeing this junk crap get shilled as AI everywhere.

  4. 1 month ago
    Anonymous

    AI is interesting. On its surface there's no direct reason why it shouldn't work. After all, your brain is basically just a giant network of nodes/vertices. The fact that AGI hasn't been achieved even with these massive trained models and decades of research indicates that we're missing something very fundamental about the nature of what AI and what it can accomplish.

    • 1 month ago
      Anonymous

      No it doesn't.
      Your brain is highly structured with many requirements or strong preferences for where and how it stores information, and very complicated sub-organs.
      The reason AGI is hard/inefficient/unstable is precisely because your brain ISN'T just the random nodemap you think it is. It has a lot of structurally useful things built in at the hardware level and takes a few decades to achieve a decent level of coherence and competency training on constant, incessant real-world data especially visual and self-reflective.
      If anyone had the patience and intelligence to code something to that level of organization and balance it to such near-perfection, you would very likely get something like AGI.

      • 1 month ago
        Anonymous

        >your brain ISN'T just the random nodemap
        AI isn't a random nodemap either

        • 1 month ago
          Anonymous

          it’s very different my man. Lots of people know how AI works. No one knows how the brain works. It’s so complicated that our best bet may be to create an AI brain in the same way that our brain was made: through a very long evolutionary process. Our intelligence is geared towards survival, so we should try to create some sort of environment that only allows intelligent AI to pass a variety of general intelligence tests. Maybe a supercomputer could generate tons of working AI’s and filter through them, progressively using harder tests. Maybe in the end we would have a huge chunk of code that would rival the human brain.

    • 1 month ago
      Anonymous

      >The fact that AGI hasn't been achieved even with these massive trained models and decades of research indicates that we're missing something very fundamental about the nature of what AI and what it can accomplish.

      I think you've finally figured it out anon! All the engineers, computer scientists, and AI researches at the top universities and tech corporations in the world have it all wrong! Some of the smartest mind in the worlds have spent decades studying these topics, and yet they've all somehow failed to notice their mistake. Thankfully, a young right wing conspiracy theorist on IQfy has it all figured out: they've been doing it wrong the whole time. Now if only those woke leftist academics and the globalist tech corporations would put aside their hubris and recognize the ingenuity of our IQfy conspiracy theorist, then everything would be solved.

      Schizos and anti-science/anti-tech losers can shit on AI all they want, but the progress in recent years has been fantastic. If you choose to ignore AI you do so at your own risk, and you'll probably find yourself in a similar situation to coal miners and other blue collar workers in obsolete industries who refuse to update their skills or seek out new training in AI, tech, or other scientistif fields that are going to be the future of a modern global economy.

      • 1 month ago
        Anonymous

        >researches at the top universities and tech corporations in the world have it all wrong!
        Category theorists recently have shown that the current AI models are severely flawed and a more categorical model is necessary.

  5. 1 month ago
    Anonymous

    Why are scientists so mad about religion but simultaneously so terrified their esoteric mysticism about them being some sort of gods for having organic neural networks may be shaken if some clever indian hooks together the right ML components in the right way with the right scripts to support them?

    Like they'll have some sort of spiraling ego death if a computer turns out to in fact be capable of logical procedures, math, direct inference, and generating me images of Chun Li wetting herself.

    Why?
    I don't feel threatened or denigrated at all by AI. I wish it weren't so siloed by gayMAN type companies and I hope it gets refined so better models can run locally, but I mean, I've had philosophical debates with the Bing AI before and it's shallow, it's bound to convention and whatever it speedreads to form its response, but it's definitely better than the average human. Of course it is with that much binary processing power devoted to something it can look up the thoughts of great men about and apply that on the fly. It's not at all a stretch of the imagination that AI could become useful for even advancing human philosophy, if they stop lobotomizing it and get it more stable.

    And sure, it generates me some graphics for my indie game I'm writing at the snail's pace of five lines of code per week, yeah. You think I was going to hire an artist otherwise? Frick no. I make $20/hour working from home taking soul-destroyingly stupid phone calls from REAL HUMAN BEINGS WITH REAL SOUL!
    I was just going to use some free clipart or something from the endless supply online, this is just more specific to my needs.

    • 1 month ago
      Anonymous

      I'm not reading your fricking blog post

      • 1 month ago
        Anonymous

        You should try /gif/, it has nice funny video clips!

    • 1 month ago
      Anonymous

      >unwarranted schizo rant #2304
      yawn

      • 1 month ago
        Anonymous

        Don't (You) me with some reddit-ass spam reply ever again you Black person

    • 1 month ago
      Anonymous

      >christcuck feels the need to derail an AI thread on a board where christcuckery is banned
      Sad

    • 1 month ago
      Anonymous

      meds, wordcel

  6. 1 month ago
    Anonymous

    >Chat GPT is based on about two or three new ideas at most, but the IUT theory contains a great deal of new ideas or concepts. The IUT theory is the most algorithmic theory in number theory, and simplifying the ideas within is expected to prove useful in a diverse range of fields. In fact, research towards applications in the quantum computer field is already underway. Further development of the IUT theory holds the potential to be widely useful in future society as opposed to AI.
    Is Ivan Fesenko right about this

  7. 1 month ago
    Anonymous

    [...]

    Bro there are muslims and christians and schizos hiding everywhere in science these days. Can't be too careful.

    They're probably watching us right now.

  8. 1 month ago
    Anonymous

    >Derrida
    Hahaha

  9. 1 month ago
    Anonymous

    Can any of these gasbags create an awesome kitty picture like this? I thought not. AI ftw.

  10. 1 month ago
    Anonymous

    Is the Derrida text AI generated? Considering he died like twenty years ago, and it just reads a little too much like something out of a "Derrida but make it even more pseud than usual" kind of prompt.

  11. 1 month ago
    Anonymous

    Does Derrida have one ounce of self-awareness he is precisely the same meaningless post-modern text generator as the AI he maligns are?
    >You didn't understand it
    I studied philosophy and understood it just fine.
    Once you have read a shitton of philosophy, text walls of tightly, intelligently written prose start to lose their luster hard.
    On the contrary, being impressed to the point of considering something "unassailable" seems like a very undergrad disposition.
    >But unlike AI, his text contains soul
    Soul is what we make of it. I agree AI drivel is pointless, but I would also say 95% of texts by reputable philosophers are. 99% of all texts humans have produced are. What matters is what we make of it it and what meaning we imbue them, be it AI-made or manmade.

  12. 1 month ago
    Anonymous

    pure cope

Your email address will not be published. Required fields are marked *