sp332

49 ms·

1. 631.▲
   by sp332 3y ago

   This might not be as in-depth as you need, but Joel Spolsky's "You Suck at Excel" is a great first hour. https://youtube.com/watch?v=0nbkaYsR94c
2. 632.▲
   by sp332 3y ago

   I don't see any code in there. What runtime could load these weights?
3. 633.▲
   by sp332 3y ago

   Hard drive prices stopped going down for a while, but now they are cheap again.
4. 634.▲
   by sp332 3y ago

   Neurodiversity is a framework for understanding human brain function and mental illness. It argues that diversity in human cognition is normal and that some conditions classified as mental disorders are differences and disabilities that ar
5. 635.▲
   by sp332 3y ago

   When a company invents tech that can do this, how would their ad be different?
6. 636.▲
   by sp332 3y ago

   They don't own the copyright, but they do have a "non-exclusive, royalty-free, transferable, sub-licensable, worldwide license to host, use, distribute, modify, run, copy, publicly perform or display, translate, and create deriv
7. 637.▲
   by sp332 3y ago

   A Voyager mission only tells us about things in our solar system. A big telescope can tell us about the rest of the universe.
8. 638.▲
   by sp332 3y ago

   https://www.earthspecies.org/ The motivating intuition was that modern machine learning can build powerful semantic representations of language which we can use to unlock communication with other species.
9. 639.▲
   by sp332 3y ago

   The Archive had a power outage tonight, do you think that might be related?
10. 640.▲
    by sp332 3y ago

    I was just copying what someone else said in this thread. It might actually be network.stricttransportsecurity.preloadlist but I have not tried it.
11. 641.▲
    by sp332 3y ago

    Firefox is fast though.
12. 642.▲
    by sp332 3y ago

    Compared to typing a nonstandard string in a nonstandard place? Going to about:config where all the deep settings are and searching for "ocsp" is downright obvious.
13. 643.▲
    by sp332 3y ago

    Peter Jackson added more motion blur in those because he said that it played better with a focus group. I think sticking to a normal 180-degree shutter angle would not feel so weird.
14. 644.▲
    by sp332 3y ago

    And don't forget overscan.
15. 645.▲
    by sp332 3y ago

    "U.S. income inequality grew through pandemic years" https://www.reuters.com/world/us/us-income-inequality-rose-3...
16. 646.▲
    by sp332 3y ago

    Isn't the US at record high levels of credit card debt right now?
17. 647.▲
    by sp332 3y ago

    I think the "What is this" link on the page just reloads the same page.
18. 648.▲
    by sp332 3y ago

    If you put it into the tokenizer https://platform.openai.com/tokenizer you can see that it helps in some places but not in others. It pulled out "SEA/OF/TRAN/QU/ILITY", but I think it broke up
19. 649.▲
    by sp332 3y ago

    You can save some space, but it still takes more than one layer to do an embedding.
20. 650.▲
    by sp332 3y ago

    We have tested this code on a 16GB Nvidia T4 GPU. The entire inference process uses less than 4GB GPU memory. You don't need 16GB of VRAM for this, it's just what they had available for testing.
21. 651.▲
    by sp332 3y ago

    There's a huge drop in the loss curve right at 8k steps. Isn't that unusual?
22. 652.▲
    by sp332 3y ago

    Are you sure it was recent and not from 1990?
23. 653.▲
    by sp332 3y ago

    That post doesn't load correctly for me in a browser. It links to https://support.google.com/photos/answer/6153599?hl=en&co=GE... and quotes this part: "Important: You can only update or remove estim
24. 654.▲
    by sp332 3y ago

    I agree with your conclusion, but it's a natural question. Twitter management and users were not in favor of him owning the place, so why not let him back out? I think the driving answer was money. He paid way more than the company was
25. 655.▲
    by sp332 3y ago

    Getting to tens of thousands of qubits is very helpful for quantum error correction schemes. That can help mitigate decoherence.
26. 656.▲
    by sp332 3y ago

    And the lack of accessibility. You generally couldn't even copy and paste text out of it.
27. 657.▲
    by sp332 3y ago

    Yeah? It was targeted for destruction by Apple because it was buggy and insecure, not because it wasn't delivering.
28. 658.▲
    by sp332 3y ago

    Yes but targeting WASM and SQLite minimizes that pain quite a bit.
29. 659.▲
    by sp332 3y ago

    Just make sure you’re buying something with a warranty. Hardware failures during training are quite common.
30. 660.▲
    by sp332 3y ago

    Is Euler Ancestral recommended for this? I thought the Ancestral samplers added noise to every step, preventing convergence.