Live data from Hacker News

Viewing profile — staticman2

staticman2

HN member
Joined
Tue, Oct 02, 2018, 10:25 PM UTC
HN karma
2,002
Public activity
1,694 items

About staticman2

No profile information was provided.

Recent public activity

  1. comment
  2. comment
    Comment #49248819

    What does this have to do with Muse Glimmer 30B?

  3. comment
  4. comment
    Comment #49210806

    If the market expected a worse job loss than 23,000 that would make sense. I'm not actually suggesting there's a good reason for market swings, just that bad news shouldn't necessa…

  5. comment
  6. comment
    Comment #49189219

    I'm pretty sure you are confusing your state's zoning laws with federal environmental regulations.

  7. comment
    Comment #49187222

    You write as if nobody has ever raised interest rates before and this is some unprecedented event that no bank could be prepared for.

  8. comment
    Comment #49180285

    Does OpenAI give Artificial Analysis early access to test models? If so it's definitely not "some random account".

  9. comment
    Comment #49143363

    Yes it tends to be correct because the other option, invent a trading scheme, tends to work out worse for investors who attempt it.

  10. comment
    Comment #49026300

    Is "You wouldn't distill..." the hot new Twitter meme or something?

  11. comment
    Comment #49025834

    3 to 5 years of cash or a bond ladder won't help in a 1970s stagflation scenario.

  12. comment
  13. comment
    Comment #48998536

    I don't agree with you on what the term SOTA means. Gemini Pro 3.1 was the only frontier model that had native video input at launch. It was certainly SOTA at some tasks.

  14. comment
  15. comment
    Comment #48993705

    It may be a dud. The blog post assures us they are now pretraining Gemini 4.

  16. comment
    Comment #48993460

    You are exaggerating. 3.1 Pro came out in February of this year.

  17. comment
    Comment #48969963

    > the idea that self-identifying as Claude is affirmative evidence that it's not distilled I wasn't arguing that. I was arguing that even if it was distilled from Claude, the disti…

  18. comment
  19. comment
    Comment #48964966

    For context you just asked me to read a document that starts with: "Caveat: fully AI-generated research." And that you quoted or paraphrased directly.

  20. comment
    Comment #48964818

    > This behavior is exactly what you'd expect from a model distilled from Claude. This is not at all what I would expect because it's trivial to change the training data to replace …

  21. comment
  22. comment
    Comment #48949051

    > If they refuse to coöperate with Apple, they’ll get personally sued as well. This isn't law and order and that's not how civil litigation works.

  23. comment
    Comment #48948998

    Would they have had to disclose a known Apple lawsuit threat in the IPO disclosures? If so that might explain the delay... Also Apple could have filed the litigation right before t…

  24. comment
    Comment #48947290

    Uber was not forced to leave the self driving car market by Waymo's litigation. The litigation ended in February 2018 and Uber left the market in December 2020.

  25. comment
    Comment #48940653

    Fair use is a defense when using copyrighted data. It is not a declaration that the data isn't copyrighted. So they are in fact literally putting copyrighted data into the model we…