jerpint

21 ms·

1. 211.▲
   by jerpint 3y ago

   Nice catch, seeing as this is implemented by pretty much every LLM out there, it’s probably not such a big deal?
2. 212.▲
   by jerpint 3y ago

   Mostly it’s just that git makes it unnecessarily hard to do atomic commits across a file
3. 213.▲
   by jerpint 3y ago

   I’m starting to wonder if the needle haystack benchmark is maybe too easy
4. 214.▲
   by jerpint 3y ago

   There are no guarantees with RAG either, and RAG only works when the answer to the question is already printed out somewhere explicitly in the text, otherwise it’s definitely prone to hallucinate
5. 215.▲
   by jerpint 3y ago

   Neat
6. 216.▲
   by jerpint 3y ago

   Looks like it’s just RAG? The paper is proposing an alternative to RAG due to its shortcomings
7. 217.▲
   by jerpint 3y ago

   The code + prompts is all open sourced so of course you could modify it to your liking, it was a fun weekend project and nothing more
8. 218.▲
   by jerpint 3y ago

   I feel like having ipython available whenever I feel like it from my shell is the best of both worlds
9. 219.▲
   by jerpint 3y ago

   I feel like you could probably generalize this method and attempt to get better performance with LLMs
10. 220.▲
    by jerpint 3y ago

    Ah didn’t know that, definitely was not possible when I made the video, but also it’s fun to hack these things together :)
11. 221.▲
    by jerpint 3y ago

    I wrote something similar using python if anyone is interested: https://www.jerpint.io/blog/yougptube/ I also made a YouTube video about it and summarized it using the tool: https://m.youtube.com/w
12. 222.▲
    by jerpint 3y ago

    This looks really promising! Other than this sentence: > We curated a large dataset of videos and languages from public book and video datasets, consisting of videos of diverse activities and long-form books. I didn’t see any other menti
13. 223.▲
    by jerpint 3y ago

    Agreed, I do this all the time especially when the model hits a dead end
14. 224.▲
    by jerpint 3y ago

    I don’t understand the obsession of LOC for wrappers - it’s the whole point of a wrapper. It makes it much easier for the user at the expense of making it less hackable Title should be instead “Library for low-code RLHF in python”
15. 225.▲
    by jerpint 3y ago

    You need >500 points to downvote
16. 226.▲
    by jerpint 3y ago

    They are not putting into question anything other the veracity of what is being shown, given that: 1) it’s being pushed by IRGC 2) the anchor is clearly a deepfake so why give undo credence to anything else the message is delivering?
17. 227.▲
    by jerpint 3y ago

    > unconfirmed Jewish people Do you even think before you type?
18. 228.▲
    by jerpint 3y ago

    An open source retrieval augmented generation (RAG) framework: https://www.github.com/jerpint/buster
19. 229.▲
    by jerpint 3y ago

    I didn’t know you could interact with pages like that so easily with Chrome extensions
20. 230.▲
    by jerpint 3y ago

    Instagram is the only point I’ll disagree on - it’s mostly become promoted content with friends content representing a fraction of what you’re being exposed to Best is to not have the app and use it through the browser to avoid mindlock
21. 231.▲
    by jerpint 3y ago

    Same here, my tmux bindings are hardwired in my brain at this point and switching over would incur a huge drop in productivity at first, not sure I have it in me to switch… floating pane does sound nice though
22. 232.▲
    by jerpint 3y ago

    I’m listening to “the creative act - Rick Rubin”, the audiobook is narrated by him and he has such a soothing voice , highly recommended
23. 233.▲
    by jerpint 3y ago

    You could probably annotate a graph , export it as an image and ask GPT4V for its best guesses, I suspect it’ll work decently well for simpler scenarios
24. 234.▲
    by jerpint 3y ago

    Is it that easy to interact programmatically with twitter? Doesn’t that make it ripe for bots?
25. 235.▲
    by jerpint 3y ago

    It’s an interesting problem to test but performed in a non-reproducible setting, so everything needs to be taken with a grain of salt
26. 236.▲
    by jerpint 3y ago

    Huggingface is slowly trying to get their own user base to adopt paper conversations
27. 237.▲
    by jerpint 3y ago

    Same! Maybe someone could do some language learning on top of all the videos we consume on YouTube?
28. 238.▲
    by jerpint 3y ago

    I don’t think the game is trying to lie to you, rather the implementation hasn’t been polished yet, no need to be so harsh
29. 239.▲
    by jerpint 3y ago

    For now RAG is the best “hack” to achieve this at very low cost since it doesn’t require any fine tuning I’ve implemented a RAG library if you’re ever interested but they are a dime a dozen now :) https://www.github.com/jerp
30. 240.▲
    by jerpint 3y ago

    They all probably already use elements of deep learning but are very likely trained in a supervised way to output structured data (I.e. actions)