411 pointsby armanified9 hours ago35 comments
  • ordinarily6 hours ago
    It's genuinely a great introduction to LLMs. I built my own awhile ago based off Milton's Paradise Lost: https://www.wvrk.org/works/milton
  • hackerman7000035 minutes ago
    Finally an LLM that's honest about its world model. "The meaning of life is food" is arguably less wrong than what you get from models 10,000x larger
  • mudkipdev3 hours ago
    This is probably a consequence of the training data being fully lowercase:

    You> hello Guppy> hi. did you bring micro pellets.

    You> HELLO Guppy> i don't know what it means but it's mine.

    • functional_dev2 hours ago
      Great find! It appears uppercase tokens are completely unknonw to the tokenizer.

      But the character still comes through in response :)

  • zwaps3 hours ago
    I like the idea, just that the examples are reproduced from the training data set.

    How does it handle unknown queries?

    • 6 minutes ago
      undefined
  • cbdevidal6 hours ago
    > you're my favorite big shape. my mouth are happy when you're here.

    Laughed loudly :-D

    • vunderba5 hours ago
      This is a direct output from the synthetic training data though - wonder if there is a bit of overfitting going on or it’s just a natural limitation of a much smaller model.
  • ankitsanghi4 hours ago
    Love it! I think it's important to understand how the tools we use (and will only increasingly use) work under the hood.
  • cpldcpuan hour ago
    Love it! Great idea for the dataset.
  • kubrador3 hours ago
    how's it handle longer context or does it start hallucinating after like 2 sentences? curious what the ceiling is before the 9M params
  • kaipereira4 hours ago
    This is so cool! I'd love to see a write-up on how made it, and what you referenced because designing neural networks always feel like a maze ;)
  • brcmthrowaway3 hours ago
    Why are there so many dead comments from new accounts?
    • loveparade2 hours ago
      It really seems it's mostly AI comments on this. Maybe this topic is attractive to all the bots.
    • AlecSchueler3 hours ago
      They all seem to be slop comments.
  • martmulx5 hours ago
    How much training data did you end up needing for the fish personality to feel coherent? Curious what the minimum viable dataset looks like for something like this.
  • gnarlouse5 hours ago
    I... wow, you made an LLM that can actually tell jokes?
    • murkt2 hours ago
      With 9M params it just repeats the joke from a training dataset.
  • NyxVox5 hours ago
    Hm, I can actually try the training on my GPU. One of the things I want to try next. Maybe a bit more complex than a fish :)
  • monksy2 hours ago
    Is this a reference from the Bobiverse?
  • rclkrtrzckr2 hours ago
    I could fork it and create TrumpLM. Not a big leap, I suppose.
  • SilentM687 hours ago
    Would have been funny if it were called "DORY" due to memory recall issues of the fish vs LLMs similar recall issues :)
  • AndrewKemendo7 hours ago
    I love these kinds of educational implementations.

    I want to really praise the (unintentional?) nod to Nagel, by limiting capabilities to representation of a fish, the user is immediately able to understand the constraints. It can only talk like a fish cause it’s very simple

    Especially compared to public models, thats a really simple correspondence to grok intuitively (small LLM > only as verbose as a fish, larger LLM > more verbose) so kudos to the author for making that simple and fun.

    • dvt6 hours ago
      > the user is immediately able to understand the constraints

      Nagel's point was quite literally the opposite[1] of this, though. We can't understand what it must "be like to be a bat" because their mental model is so fundamentally different than ours. So using all the human language tokens in the world can't get us to truly understand what it's like to be a bat, or a guppy, or whatever. In fact, Nagel's point is arguably even stronger: there's no possible mental mapping between the experience of a bat and the experience of a human.

      [1] https://www.sas.upenn.edu/~cavitch/pdf-library/Nagel_Bat.pdf

      • Terr_3 hours ago
        IMO we're a step before that: We don't even have a real fish involved, we have a character that is fictionally a fish.

        In LLM-discussions, obviously-fictional characters can be useful for this, like if someone builds a "Chat with Count Dracula" app. To truly believe that a typical "AI" is some entity that "wants to be helpful" is just as mistaken as believing the same architecture creates an entity that "feels the dark thirst for the blood of the living."

        Or, in this case, that it really enjoys food-pellets.

      • andoando2 hours ago
        Id highly disagree with that. Were all living in the same shared universe, and underlying every intelligence must be precisely an understanding of events happening in this space-time.
      • 4 hours ago
        undefined
      • AndrewKemendo6 hours ago
        Different argument

        I’m not going to argue other than to say that you need to view the point from a third party perspective evaluating “fish” vs “more verbose thing,” such that the composition is the determinant of the complexity of interaction (which has a unique qualia per nagel)

        Hence why it’s a “unintentional nod” not an instantiation

        • 4 hours ago
          undefined
  • nullbyte8087 hours ago
    Adorable! Maybe a personality that speaks in emojis?
  • oyebenny4 hours ago
    Neat!
  • dinkumthinkum4 hours ago
    I think this is a nice project because it is end to end and serves its goal well. Good job! It's a good example how someone might do something similar for a specific purpose. There are other visualizers that explain different aspects of LLMs but this is a good applied example.
  • 6 hours ago
    undefined
  • peifeng07an hour ago
    [dead]
  • zephyrwhimsyan hour ago
    [dead]
  • techpulselab38 minutes ago
    [dead]
  • Alexzoofficial3 hours ago
    [flagged]
  • Morpheus_Matrix7 hours ago
    [flagged]
  • ethanmacavoy5 hours ago
    [flagged]
  • agenexus5 hours ago
    [flagged]
  • george_belskyan hour ago
    [dead]
  • aesopturtle4 hours ago
    [flagged]
  • weiyong10246 hours ago
    [flagged]
  • aditya73030114 hours ago
    [dead]
  • aditya73030114 hours ago
    [dead]
  • jiusanzhou2 hours ago
    [flagged]
    • ngruhn2 hours ago
      comment smells AI written
    • 3m2 hours ago
      AI account