Comments by "clray123" (@clray123) on "Freya Holmér" channel.

  1.  @SahilP2648  Not really unless you're talking about boilerplate. Otherwise the LLM is good to get you started and give some clues what to research, but trying to make it actually write working code is an exercise in endless frustration, with it spinning in circles adding new bugs and slop while trying to "fix" the old ones you tell it about.
    31
  2.  @SahilP2648  lol cut-and-paste and generic website customizers/generators existed before, you will have to make a better effort to come up with impressive examples because the reality is any time you try to step out of what was already baked in during training the ai generators fail miserably
    18
  3. The funny thing is that Google's AI is not even capable enough to detect and ban AI-generated content.
    11
  4.  @AnarchistArtificer  LLMs (and generative models in general) are bad at anything other than some random interpolation and variation of the same content they've been fed in training data - which is enough for some "artistic" endeavors if the "art" boils down to regurgitating old content (which admittedly a lot of commercial art boils down to). If you have anything non-trivial to solve through that same random guessing process, you have to spend hours of computation time and thousands of dollars in hardware costs (as OpenAI's newest and "most impressive" o3 model recently confirmed). In other words the algorithms are entirely and economically unsuitable for the tasks that they are advertised to perform.
    7
  5.  @NLPprompter  AI would be ok term if the "I" stood for "Imitation" (or "Interpolation").
    6
  6. That's a great question!
    5
  7. AI - the ultimate imposter/pretender.
    3
  8. You should blame AI slop on those who use AI tools in sloppy way. Blaming the tool is as if I was blaming cameras for films made by a shitty director.
    3
  9.  @jennysquibb7440  This is exactly what will happen, and in a way, it will make real above-average skills more valuable - because all the mediocre slop will also reduce incentives to obtain such skills. Humanity overall will become dumber, cheaper, less happy, but some/few will thrive because of it.
    2
  10. The weird wording "graphical games" and grammar "open file extension" does not feel like AI, it feels like Indian English. It is probably cheap Indian laborers using AI to generate these websites. State-of-the-art AI writes much better than what you see there.
    2
  11. Instead of searching Google for Ai generated websites why not just ask Claude AI to generate all the code for you.
    1
  12. BTW this destruction of the Internet is not a coincidence, it is a ploy by people like Sam Altman to push a universal digital ID system which will "solve" the self-created problem by allowing you to "prove" that you are human and also to "certify" human content. The overall goal is not anything less than unviersal continuous surveillance of every human being on the planet. If you can tell who someone is, what they think, and can control virtually all their communications with others, then you have the ultimate power over that individual. And when you are a half-crazy sociopath narcissist, obtaining ultimate power over (all) other humans feel very enticing.
    1
  13. Anytime someone says "That's a great question" during a conversation, red alarm lights are blinking in my head - either the individual is an American idiot or an AI bot.
    1
  14.  @fluffycritter  The best thing is the quotations where it quotes random irrelevant pieces of the source text while ignoring the actually important ones. Sometimes it quotes a piece that directly contradicts what it says.
    1
  15. The common snowflake experience of growing up in a real world. If you are raised in harsher conditions, you don't have any such fairy tale illusions of what "should be", hence no "depression".
    1
  16.  @XeZrunner  When it comes to AI slop, we're talking about one-"business"man run sweatshops in India, not real companies. These people are simpy desperate for money, their business is a bit like running a digital prostitution racket. They also don't really care about their "customers", they care about somehow exploiting the system to generate a small profit in a dog-eat-dog world which surrounds them (because of overpopulation and global politics).
    1
  17. There's more to it, they want to eliminate your privacy and sell you "prove that you're human online" technology (universal digital ID and access control). Utlimately it is all about permissions managment and military use. That's why there's so much money in it, supported by biggest giant corporations.
    1
  18. Nah, you just need stronger AI which will filter out inferior AI crap. P.S. If you feed the text of such a website to ChatGPT or Claude today, it will likely be able to highlight all the mistakes.
    1
  19.  @funicon3689  nothing, an increasingly better AI will be the result
    1