• 0 Posts
  • 22 Comments
Joined 1 year ago
cake
Cake day: July 4th, 2023

help-circle






  • Haha that reminds me of the time I helped my upstairs neighbors move. They were super annoying, always shouting over each other and leaving trash bags in the hot sun on their balcony, where the stink could waft into my open window.

    I was very excited when I saw them packing their furniture, so I offed to help and we managed to finish it all in only one day. They didn’t catch on and just thought I was being nice! They even tried to pay me but I refused - being rid of them was payment enough





  • It’s also about making sure you can’t sue them, even if they did something wrong, even if they did it on purpose, even if they knew it was wrong when they did it.

    Instead you must agree to “binding arbitration”, so that if you lose they get to learn what strategy works against customers, and if you win they get to learn what strategy doesn’t work against customers (but in any case the details cannot be shared with other customers).

    Also, you can never participate in a class action suit, so even if they did do something wrong, on purpose, and you convinced a -judge- arbiter, you just get the $12 judgment, or whatever the value is of your actual damages. The corporation can keep the $12 they stole from each of their millions of other customers, who didn’t also start arbitration.

    It is shocking that it is even legal






  • I think I understand how it works.

    Remember that LLMs are glorified auto-complete. They just spit out the most likely word that follows the previous words (literally just like how your phone keyboard suggestions work, just with a lot more computation).

    They have a limit to how far back they can remember. For ChatGPT 3.5 I believe it’s 24,000 tokens.

    So it tries to follow instruction and spits out “poem poem poem” until all the data is just the word “poem”, then it doesn’t have enough memory to remember its instructions.

    “Poem poem poem” is useless data so it doesn’t have anything to go off of, so it just outputs words that go together.

    LLMs don’t record data in the same way a computer file is stored, but absent other information may “remember” that the most likely word to follow the previous word is something that it has seen before, i.e. its training data. It is somewhat surprising that it is not just junk. It seems to be real text (such as bible verses).

    If I am correct then I’m surprised OpenAI didn’t fix if. I would think they could make it so in the event the LLM is running out of memory it would keep the input and simply abort operation, or at least drop the beginning of its output.