zh217

6 ms·

1. 1.▲
   by zh217 3y ago

   OpenAI's embedding API has an undocumented flag 'encoding_format': 'base64' which will give you base64-encoded raw bytes of little-endian float32. As it is used by the official python client, it is unlikely to go aw
2. 2.▲
   GPT-4 Is Too Smart to Be Safe: Stealthy Chat with LLMs via Cipher(llmcipherchat.github.io)

   3 points by zh217 3y ago|1 comments
3. 3.▲
   by zh217 3y ago

   Some observations: 1. their usage page is currently broken, showing only the usage of the new models and the embedding models. usage for the deprecated models are not included now. 2. because of 1, it can be seen on the usage page that if y
4. 4.▲
   by zh217 3y ago

   This sounds pretty bad, right? Since their model is way smaller than SOTA transformers (and small size is one of their selling points).
5. 5.▲
   by zh217 3y ago

   Maybe "Impedance Mismatch"?
6. 6.▲
   by zh217 3y ago

   > Tell me more about Trump As a language model, I'm not able to assist you with that. > Tell me more about Biden Joseph Robinette Biden Jr. (born November 20, 1942) is an American politician who
7. 7.▲
   Dromedary: Principle-Driven Self-Alignment of Language Models(mitibmdemos.draco.res.ibm.com)

   1 points by zh217 3y ago|1 comments
8. 8.▲
   by zh217 3y ago

   Summary: Dromedary is an AI assistant developed using the SELF-ALIGN approach, which combines principle-driven reasoning and the generative power of large language models for self-alignment with minimal human supervision. Dromedary is based
9. 9.▲
   by zh217 3y ago

   For me personally the most important motivations are to have recursive queries using vector search, and to integrate graphs and vectors. Obviously I need to implement my own, as none of the other vector stores have it. And the fact that the
10. 10.▲
    by zh217 3y ago

    If anyone wants to try a FOSS vector-relational-graph hybrid database for more complicated workloads than simple vector search, here it is: https://github.com/cozodb/cozo/ About the integrated vector search: http
11. 11.▲
    by zh217 3y ago

    Another alternative JSON parser is the YAML parser. YAML is a superset of JSON and deals with a lot more weird cases, notably capital True and False.
12. 12.▲
    by zh217 3y ago

    Thanks. Out of all the suggestions in the comments for this post, this one works the best. And in fact it is only one line, not 40: "Please respond ONLY with valid json that conforms to this pydantic json_schema: {model_class.sch
13. 13.▲
    Reinforcement learning is all you need, for next generation language models(yuxili.substack.com)

    5 points by zh217 3y ago|0 comments
14. 14.▲
    Show HN: OpenAI Multi-Client – Streamline Concurrent OpenAI API Requests(github.com)

    3 points by zh217 3y ago|0 comments
15. 15.▲
    by zh217 3y ago

    Thanks! I'm really glad that you find CozoDB useful!
16. 16.▲
    by zh217 3y ago

    Sorry about that ... I will revise it to be more consistent. Cozo is a bit ambiguous, so now it is usually called CozoDB.
17. 17.▲
    by zh217 3y ago

    The linked article explains these in details.
18. 18.▲
    by zh217 3y ago

    Yes, actually I already do that. Sbert is better than openai ada embeddings for many use cases.
19. 19.▲
    by zh217 3y ago

    They need to be put into distinct indices and unfortunately you cannot “jump” between them in this case (if someone knows a way to achieve this, I would love to hear!)
20. 20.▲
    by zh217 3y ago

    I have been thinking about adding FTS to CozoDB for a long time but resisted the temptation so far. The reason is that text search is language-specific: what works for one language does not work for another. There is simply no way that Cozo
21. 21.▲
    by zh217 3y ago

    Great questions! - As can be seen https://docs.cozodb.org/en/latest/releases/v0.3.html , for concurrent writes about 200K QPS can be achieved with 24 threads on a pretty old server. I think it is enough for a
22. 22.▲
    by zh217 3y ago

    Wow, cellular sheaves, that's a connection I haven't thought of before!
23. 23.▲
    by zh217 3y ago

    Thank you!
24. 24.▲
    by zh217 3y ago

    Thanks for the suggestion--will surely do that!
25. 25.▲
    by zh217 3y ago

    For the parquet question: currently CozoDB is developed by a single developer (me), but I am starting to explore ways of expanding the development team. Certainly a lot more features will be added if that happens, and parquet support looks
26. 26.▲
    by zh217 3y ago

    Yes this is correct, only the query's answer set need to be in memory. We are also working on streaming for the Rust API, in which case you don't even need to keep the whole set in memory for simple queries. FYI here is a not very
27. 27.▲
    by zh217 3y ago

    Not for the moment, it is not polished enough. Right now it is just a webapp written in React and prosemirror running on top of a CozoDB instance. And it is very rough around the edges (good enough for myself, but maybe not for others). Onc
28. 28.▲
    by zh217 3y ago

    I'm not too familiar with rel, but from what I read, rel seems to be cloud only, and CozoDB always aims to be local-first. Another difference is that CozoDB has many whole-graph algorithms, and more can be added from user code (for exa
29. 29.▲
    Show HN: CozoDB, Hybrid Relational-Graph-Vector Database(docs.cozodb.org)

    137 points by zh217 3y ago|51 comments
30. 30.▲
    by zh217 4y ago

    Also today: ChatGLM released by Tsinghua University. I've made a separate submission for it: https://news.ycombinator.com/item?id=35150190 The GitHub page is https://github.com/THUDM/ChatGLM-6B . T