Comments by "Charles M." (@charlesm.2604) on "GPT-4.5 shocks the world with its lack of intelligence..." video.

  1. LLMs are glorified guesstimate algorithms
    17
  2.  @XDarkGreyX  Because my washing machine is programmable too. But to avoid confusion, I usually describe my job as "software developer", it clears the air between the dummies who would think someone spends his days pushing washing machine buttons and the dummies who are too purists to bend the rules for the sake of readability like yourself.
    10
  3. ​@ I am not saying they aren't productive. LLMs are literally glorified guesstimate algorithms. It's how they work. I wasn't downplaying how good their results are, just that there is nothing to be done to make them better. Sanitizing datasets, sure. Feeding them more data, increasing its parameters (until it LITERALLY can't fit on the planet), sure. Using more efficient transformers, sure. That's the extent of the progress of LLMs, it LITERALLY can't be better than it is. I can't wait for the LLM hype to be over and for the research to be done on actual AI. We have made so much advancement in computer vision, but we somehow abandoned it. In a year we'll have companies training their LLMs distills using their own data for their own specific targeted needs (without the insane general purpose weights of the massive models like OAI, Claude, LLama, etc.) and we will move back to fund AI development in directions that'll actually matter. I encourage you to read more on the topic of what LLMs are, and how they work. In the meantime, here's the takeaway from Google's own "Introduction to Large Language Models" papers: "Language models predict and generate text by estimating the probability of a token or sequence of tokens occurring within a longer sequence, useful for tasks like text generation and translation. Large language models (LLMs) are advanced language models with vast parameters and datasets, enabling them to process longer text sequences and perform complex tasks like summarization and question answering." This reply is probably going to get deleted but I hope you can read it before it happens.
    1
  4.  @Brandon82967  What do you think the difference between o1 and o3 is? If you took my advice and learned about how LLMs work instead of throwing yourself at your keyboard you'd know the answer.
    1