ukuina

39 ms·

1. 661.▲
   by ukuina 3y ago

   I would recommend AI Applications Engineering (e.g., applying LLMs as a unit of compute). Start with reading AI Canon and setting up projects like PrivateGPT and AutoGPT locally, then working with LocalAI to serve up HuggingFace models in p
2. 662.▲
   by ukuina 3y ago

   Anyone building vectorDBs in-browser, possibly using WASM?
3. 663.▲
   by ukuina 3y ago

   Are you saying the lines are blurred, or that there are more tiers of abstraction?
4. 664.▲
   by ukuina 3y ago

   They have also replaced previously metal joints and screws with wooden pieces. I had to bolster my chest of drawers with additional wooden center beams to prevent it from bowing from evenly distributed weight. Oddly, there were already prov
5. 665.▲
   by ukuina 3y ago

   "Zero AI" will likely become a selling point.
6. 666.▲
   by ukuina 3y ago

   Especially given GPT-based news summarization services.
7. 667.▲
   by ukuina 3y ago

   I would argue there are four prices: acquisition, usage, storage, and disposal. The middle two are (ideally?) recurrent costs and storage costs in the age of consumerism grow linearly throughout life. At some point, people "see the lig
8. 668.▲
   by ukuina 3y ago

   A more grounded pitch for Hololens 2: https://youtu.be/d3YT8j0yYl0
9. 669.▲
   by ukuina 3y ago

   Incredible future VR use-case for all the hours of useless video footage in my archives.
10. 670.▲
    PandaGPT: One Model to Instruction-Follow Them All(panda-gpt.github.io)

    4 points by ukuina 3y ago|0 comments
11. 671.▲
    by ukuina 3y ago

    Perhaps we need a new term... "Armchair LLMer"?
12. 672.▲
    by ukuina 3y ago

    Doesn't compensation follow good/fast work, not precede it?
13. 673.▲
    by ukuina 3y ago

    You can use https://localai.io if you have a GPU or Apple Silicon CPU to serve up local models with an OpenAI-compatible API.
14. 674.▲
    by ukuina 3y ago

    What would that accomplish?
15. 675.▲
    by ukuina 3y ago

    Has Apple ever released a Developer Edition product before?
16. 676.▲
    by ukuina 3y ago

    This doesn't hold true for FPS games. You absolutely need to tell the game where you are aiming so it can draw crosshairs, even for on-rails shooters favored by VR.
17. 677.▲
    by ukuina 3y ago

    Sounds like the SQLite model, which has been a net positive for the computing world.
18. 678.▲
    by ukuina 3y ago

    That's a great idea, I will look at FAISS for now. Thank you.
19. 679.▲
    by ukuina 3y ago

    Thank you for the pointer. This still requires a backend server running Chroma, correct?
20. 680.▲
    by ukuina 3y ago

    Thank you for responding here. For this enterprise use-case, I am looking to ship a web app that will run off of IT-managed fileservers and IT-managed LLM APIs, on user desktops and laptops without privileged access to install software.
21. 681.▲
    Ask HN: In-Browser Vector DB?

    2 points by ukuina 3y ago|6 comments
22. 682.▲
    by ukuina 3y ago

    All of it. Globally.
23. 683.▲
    by ukuina 3y ago

    Is this effectively a watch() on a variable?
24. 684.▲
    by ukuina 3y ago

    Microsoft Office Copilot is supposed to handle this specific task, among many others, with ease.
25. 685.▲
    by ukuina 3y ago

    CPU idle % over a rolling time window (say, 15 minutes).
26. 686.▲
    by ukuina 3y ago

    The 1T number is wild hype. It's likely GPT4 is actually a mixture-of-experts with smaller models under the hood.
27. 687.▲
    by ukuina 3y ago

    [2014] The original GAN paper, likely still relevant when integrating LLMs.
28. 688.▲
    by ukuina 3y ago

    I would recommend access over API and using an interface like TypingMind that gives you control over the system prompt for consistency.
29. 689.▲
    by ukuina 3y ago

    Isn't that still on waitlist-preview?
30. 690.▲
    by ukuina 3y ago

    There's no real benefit to streaming if you are planning to use the LLM output downstream (say, in a SQL query). LLM latency is a major annoyance right now, whether locally-hosted or cloud-based.