saliagato

8 ms·

1. 31.▲
   by saliagato 3y ago

   Hopefully they don’t
2. 32.▲
   by saliagato 3y ago

   We all know what. HN moderators are deleting all related comments. Edit: dang is right, sorry y’all
3. 33.▲
   by saliagato 3y ago

   I think they are running their RAG on Microsoft's Azure Cognitive Search. Satya mentioned something about it in the yesterday Microsoft Ignite keynote
4. 34.▲
   by saliagato 3y ago

   It can’t count because what the LLM sees is a bunch of tokens not words
5. 35.▲
   by saliagato 3y ago

   Yeah same. I switched to Azure a while back but it's unusable now. Switched back to OAI API
6. 36.▲
   by saliagato 3y ago

   Basically they take your data for free and they will receive your money from the API usage you will make. Good
7. 37.▲
   by saliagato 3y ago

   You can now [1] pay from $2 to $3 million to pretrain custom gpt-n model. This has gone unnoticed but seems really neat. Provided that a start-up has enough money spend on that, it would certainly give competitive advantage. [1] https:&#x2
8. 38.▲
   by saliagato 3y ago

   I think your startup just died
9. 39.▲
   by saliagato 3y ago

   sama said they improved it
10. 40.▲
    by saliagato 3y ago

    OpenAI wants to make ChatGPT an ecosystem similar to the App Store. They are creating a diversification in their income: -Subscriptions -API -This new marketplace That’s a nice strategy
11. 41.▲
    OpenAI’s November 6th DevDay has been leaked(the-decoder.com)

    5 points by saliagato 3y ago|3 comments
12. 42.▲
    by saliagato 3y ago

    Care to elaborate?
13. 43.▲
    Ask HN: Is OpenAI exposing user data via Google Search?

    3 points by saliagato 3y ago|3 comments
14. 44.▲
    Ask HN: How does GPT-4V work?

    4 points by saliagato 3y ago|0 comments
15. 45.▲
    by saliagato 3y ago

    which?
16. 46.▲
    by saliagato 3y ago

    yes
17. 47.▲
    by saliagato 3y ago

    OP was referring to nazis
18. 48.▲
    by saliagato 3y ago

    Europe
19. 49.▲
    LLMs trained on “A is B” fail to learn “B is A”(paperswithcode.com)

    120 points by saliagato 3y ago|158 comments
20. 50.▲
    by saliagato 3y ago

    I can get 8k + tokens response from OpenAI API without problems (with gpt-3.5-turbo-16k) but I encounter errors from Azure OpenAI when using the same prompt and parameters. Obviously it sucks because I can't get long response from gpt-
21. 51.▲
    by saliagato 3y ago

    Wow. This is the future. I am going to launch a product based on this
22. 52.▲
    by saliagato 3y ago

    That’s exactly what he said
23. 53.▲
    by saliagato 3y ago

    Azure Cognitive Search takes care of all of this combining semantic search with other layers of traditional search methods
24. 54.▲
    by saliagato 3y ago

    I still find ChatGPT (4) far superior to Claude.
25. 55.▲
    Claude Pro(anthropic.com)

    1 points by saliagato 3y ago|3 comments
26. 56.▲
    by saliagato 3y ago

    Very good keep going. I am going to subscribe
27. 57.▲
    by saliagato 3y ago

    I wouldn’t trust a fyi tld
28. 58.▲
    by saliagato 3y ago

    everything but 32k version and 2x speed is the same as the consumer platform
29. 59.▲
    by saliagato 3y ago

    I thought that Microsoft was busy with enterprises yet OpenAI announces a product for enterprises. I have a feeling that the two do not get along
30. 60.▲
    by saliagato 3y ago

    Not OP but, I finally got access to GPT-4 32k and I thought that for a thing as simple as a summary it would be better than the smaller context-window model. But then I realized that now I have (practically) infinite context in input but wh