• 0 Posts
  • 17 Comments
Joined 1 year ago
cake
Cake day: June 4th, 2025

help-circle

  • AI slop is an order of magnitude worse.

    Now, all someone needs to do is give a four-sentence prompt, and it can be expanded into a twenty-page essay. It is essentially an attack on human attention… it still takes you twenty minutes to read the essay that only really is about that four-sentence prompt. The information density is intentionally diluted to ridiculous levels.

    At least someone had to make the effort to write those cheap texts.





  • Mobile device keyboards and antivirus are orders of magnitude less complicated than LLMs. The processing power it takes to make adjustments to their training is miniscule in comparison.

    How you poison LLMs is by poisoning their training data - a.k.a. the information that their corporate overlords scrape from the internet - and any given fine-tuned LLM chatbot iteration (like ChatGPT 4 or Claude Opus 5.5) is essentially locked in place upon creation. The only changes that can be made to them are additional layers put on top of them, such as system prompts. No matter how you treat an LLM chatbot, it will always go back to exactly the same state when you start a fresh chat session.




  • SparroHawc@lemmy.ziptoMicroblog Memes@lemmy.world•Yeet
    link
    fedilink
    English
    arrow-up
    2
    ·
    2 months ago

    If the tiny person wasn’t going so fast that it’s in orbit around the big thing, yeah - all you have to do to get the person to fall in is to stop them moving. The problem is that you have to push them reeeeeal hard to get them to stop moving in orbit around the sun. (That’s what Delta-v means - change in velocity.)








  • Students with access to AI are already over-relying on it. It is a dangerous tool for children to have unfettered access to, just like social media. It’s sold, misleadingly, as a panacea for having to do research yourself, having to write things yourself, having to think for yourself.

    Sure, learning how to use AI and when to use AI is useful - but its actual usefulness is far, far below the level that people who say things like “Students that embrace AI will outperform traditional students” believe it to be. Everything an AI spits out HAS to be double-checked. AI corps don’t want you to think about that aspect. Unless this is hammered home in every class that AI is used in, then it shouldn’t be used in school.

    Let me present an example from OP’s image. Julia complains that AI can write essays better than the teachers who grade the essays students write. Leaving aside the extraordinary skepticism I feel about this statement, Julia is implying that writing essays in school is a waste of time - but learning how to write a functional, persuasive essay, with proper research, sources, etc. is very, very important, because it teaches you to extract the facts you need from sources and cite them to prove your point in a way that can’t be refuted offhand. It teaches you to examine your premise, find what makes it tick, and take it apart to find the constituent parts so you can shore them up with critical details that prove it. English courses teach you about grammar, spelling, tone, context, tense… I could go on.

    Prompting an LLM to write the essay for you teaches you none of this, and claiming that the two are equivalent is the height of folly. In fact, if someone ‘writes’ an essay with an LLM, I am much more likely to simply reject it out of hand, because it means they didn’t want to take the time to make their own argument. I would rather see the prompt that they used rather than the essay itself, because there will be, in essence, zero additional information in the essay - and asking me to read LLM output as if it provided value of some sort is an insult. I can ask ChatGPT to spew out details on a topic just as well as you can. At least if someone wrote the essay themselves - even if it is poorly written - there’s the possibility of them having a novel approach to the topic.


  • Students that embrace AI might outperform traditional students in situations where AI performs well. However, they will perform worse in all other metrics, and significantly worse when they don’t have access to AI.

    This isn’t theory. Teachers are already seeing it.

    The point of school isn’t to do work. The point is to learn, and typing prompts into genAI isn’t learning.