• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI
    Reaction

    Suggested ways to make language model outputs easier to understand

    A post recommends controlled-language writing, diagrams and HTML web pages as ways to make model outputs easier to follow.

    AK
    YJ
    EL
    5 Sources, ,

    TLDR

    A user expects people to spend more time overseeing and understanding language model outputs as the models improve. They suggest asking for writing in ASD-STE100, a controlled language developed for aerospace maintenance documentation, or requesting diagrams and HTML web pages. They are especially optimistic about custom explainer videos, which they say are starting to work.

    Combined views

    506.6K

    5 Sources, first seen 4h ago

    Combined views

    506.6K

    5 Sources, first seen 4h ago

    11.7K likes
    4h ago
    first seen 4h ago
    11.7K likes
    430 comments
    13K saves
    1.1K reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Featured Source
    430 comments
    13K saves
    1.1K reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Today's Rank

    #15

    Today's Rank

    #15

    5 Sources

    @karpathyWe'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks: Writing. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent. But even better: Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better: Web pages. Ask for output "in HTML" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better: Explainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like "Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work! In summary: - As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding. - Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.
    @Yuchenj_UWI really wish I could just prompt an AI to generate an Andrej Karpathy video. Unfortunately, no AI today can do that. Andrej, we need you back. It’s been over a year since you uploaded a video to youtube.
    @Grady_Booch@karpathy "Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand." Hm, there's this thing called the UML...
    @eliebakouchyeah i'm working on "autoresearch" and i'm sooo bottlenecked by this, but opus 5.5 is an amazing model here for helping me understand the output of the different model (please continue improving it in this direction 🙏) i need to try video but i'd say that interactive HTML is probably one of the most time-efficient ways to understand model output, the only issue is that models are not good enough at building the right level of abstraction around the plots. a few example bellow that took a good number of iterations with opus but i'm still not satisfied with the level of detail and the clearness of the explanation
    @omarsar0@karpathy Yup.

    5 Sources

    @karpathyWe'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks: Writing. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent. But even better: Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better: Web pages. Ask for output "in HTML" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better: Explainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like "Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work! In summary: - As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding. - Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.
    @Yuchenj_UWI really wish I could just prompt an AI to generate an Andrej Karpathy video. Unfortunately, no AI today can do that. Andrej, we need you back. It’s been over a year since you uploaded a video to youtube.
    @Grady_Booch@karpathy "Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand." Hm, there's this thing called the UML...
    @eliebakouchyeah i'm working on "autoresearch" and i'm sooo bottlenecked by this, but opus 5.5 is an amazing model here for helping me understand the output of the different model (please continue improving it in this direction 🙏) i need to try video but i'd say that interactive HTML is probably one of the most time-efficient ways to understand model output, the only issue is that models are not good enough at building the right level of abstraction around the plots. a few example bellow that took a good number of iterations with opus but i'm still not satisfied with the level of detail and the clearness of the explanation
    @omarsar0@karpathy Yup.