rolisz

14 ms·

1. 271.▲
   by rolisz 3y ago

   Cool tool, I used it for a scrapinh project and it performed quite well for extracting clean text and the date.
2. 272.▲
   by rolisz 3y ago

   And civilization is a distributed operating system! (I heard this from Joscha Bach)
3. 273.▲
   by rolisz 3y ago

   They're also cancelling Parks and Recreation? :(
4. 274.▲
   by rolisz 3y ago

   Hah, we wish. In practice what will happen is we'll get the black hole spacetime disruption due to NPM package weight combined with dependency hell from pip
5. 275.▲
   by rolisz 3y ago

   So you can now shoot yourself in the foot with both Python and JavaScript package management?
6. 276.▲
   by rolisz 3y ago

   I've just been to two endocrinologists who gave me completely different diagnosis (they said different organs are the issue) Is this a sign endocrinology has no scientific rigor? There is another explanation for the variation in therap
7. 277.▲
   by rolisz 3y ago

   I think very, very, very few people still care about the C89 compiler. A bigger blocker to a merge/collaboration would be the move to use Lua in neovim
8. 278.▲
   by rolisz 3y ago

   Hard disagree. Evaluation is hard. Handling stupid mistakes from LLMs is something that has to be taken into account from day 1. Finetuning is not that needed in my experience.
9. 279.▲
   by rolisz 3y ago

   1) Dropbox can be replicated in a couple of hours with SFTP (or whatever was that iconic HN comment) 2) the devil is in the details. How do you get the data out of documents? Are they pdfs? Do they have tables? Do they have images? Sure, cr
10. 280.▲
    by rolisz 3y ago

    For what it's worth, LLM powered chatbots are quite different from the chatbots that were popular 5 years ago and often feel much more natural.
11. 281.▲
    by rolisz 3y ago

    I expect that for Linux or PostgreSQL it was trained not just on the codebase, but also on mailing lists and tons of other documents that discuss various things about it and that's what give it these reasoning powers.
12. 282.▲
    by rolisz 3y ago

    But finetuning wouldn't give the LLM a holistic understanding. LLMs are finetuned to predict the next token. At best you give them a full file at a time, but more likely, you create batches of similar length sequences. So while finetun
13. 283.▲
    by rolisz 3y ago

    > when you can guarantee a valid JSON syntax by only sampling tokens that are valid? Instead of greedily picking the highest-scoring token every time, you select the highest-scoring token that conforms to the requested format. Yes, you c
14. 284.▲
    by rolisz 3y ago

    Everyone. All the alternatives have some big disadvantages, so more competition is good.
15. 285.▲
    by rolisz 3y ago

    But you wouldn't have permissions to release a new version (which is what is being asked).
16. 286.▲
    by rolisz 3y ago

    sigh why are Hungarians so special? It created confusion in my mind as to which one is the first mind (I'm half Hungarian)
17. 287.▲
    by rolisz 3y ago

    I was named Roland because it exists in both Romanian (my mother's family) and Hungarian (dad's family). It's not pronounced exactly the same (the a is different), but it's easy enough for all grandparents. And I even le
18. 288.▲
    by rolisz 3y ago

    The EU is composed of 27 countries, with varying construction standards. In Romania, you can hear when your neighbors flush the toilet, especially in new constructions.
19. 289.▲
    by rolisz 3y ago

    Juniors won't be able to properly evaluate the AI generated code, so they won't be able to benefit from it fully. A senior however who can crank out 10x code will be more valuable.
20. 290.▲
    by rolisz 3y ago

    Nope, it's not guaranteed. They warn you in the OpenAI docs that it might hallucinate inexistent parameters.
21. 291.▲
    by rolisz 3y ago

    Privatization of trash collection has already happened in large parts of Romania. Sometimes the results are good, sometimes bad, but it's always much more expensive.
22. 292.▲
    by rolisz 3y ago

    The T in ChatGPT stands for Transformers. The similarity between the OG Transformer from 2017 and GPT3 (and other modern LLMs) is pretty big
23. 293.▲
    by rolisz 3y ago

    There are open source models that can deliver pretty well for chatbot over internal documentation. If you're interested, feel free to reach out to me.
24. 294.▲
    by rolisz 3y ago

    Two weeks ago I finished a project for a client who wanted a "talk to your documents" application, without using OpenAI or other 3rd party APIs, but by using open source models running on their own infrastructure. If you're i
25. 295.▲
    by rolisz 3y ago

    If you've never coded or used Python before, yeah, go with OpenAI. Otherwise, generating embeddings with SentenceBERT takes 5 minutes. And from my personal experience Ada embeddings are not the best. They are large (makes aproximate se
26. 296.▲
    by rolisz 3y ago

    FastChat-T5 can work for such a use case and it runs on (beefy) CPUs. With a 700$/month instance, it can do 4 conversations simultaneously, without needing GPUs. The instant a company has sensitive data, this becomes very viable.
27. 297.▲
    by rolisz 3y ago

    I disagree that single threadedness is in its DNA. It's an implementation detail of CPython. There are other implementations which don't have a GIL even today. Would removing the GIL be a big change for CPython? Yes. But IMO it&#x
28. 298.▲
    by rolisz 3y ago

    Moving from 2 to 3 was a long and difficult migration, so 3 to 4 will be similarly difficult
29. 299.▲
    by rolisz 3y ago

    But they do change things in the ChatGPT web app. You can't choose the exact model there, just 3/4 and from time to time they update the models they use.
30. 300.▲
    by rolisz 3y ago

    Folks using it in production are using the API, where you can explicitly select a model. In the post, the poster is using ChatGPT, through the web app, where you can't select exact model and they sometimes do updates.