korrents

korrents · video

Another win for China & Open Weight models

NeetCode · 15m · youtube.com ↗

15 korrents from this recording

NeetCode did not write this page.

Every claim below is a statement made in this recording, quoted word for word and linked to the second it was said, so you can hear it rather than take our word for it. The wording comes from the transcript published alongside the recording; the sentence above each quote is our reading of the claim, not their wording.

  1. 0:47 · watch on youtube.com↗

    it's clear that the main motivator at this point is money, which is fine, but it's also becoming clear that owning the models is like owning the means of production, and people are catching on to that.
  2. 1 min later
  3. 1:20 · watch on youtube.com↗

    The best open weight models these days are coming from China, and I think what OpenAI is doing by setting a precedent, they can just choose who has access to their models or not, is clearly not going to help this trend reverse at all.
  4. 1 min later
  5. 2:18 · watch on youtube.com↗

    Whatever any of these companies stands for or say they stand for, they will always end up contradicting themselves. And the reason that's important is because it kind of makes it hard to trust these companies when you don't really know how they're going to behave
  6. 1 min later
  7. 2:53 · watch on youtube.com↗

    It's not that these companies just wake up one day and feel like doing something different. It's all about strategy. It's all about winning the game and concentrating as much power as they can. But the way I see it going right now, long-term, it's starting to look like a losing game for American companies.
  8. 3 min later
  9. 6:04 · watch on youtube.com↗

    I think that's part of the reason that even the Chinese companies have been able to go as fast as they can because they are doing a lot of distilling on Anthropic models and probably also the open AI models as well.
  10. 6:23 · watch on youtube.com↗

    I think currently even though open AI has some very capable models, I don't think they have anything that is at the level of Fable.
  11. 2 min later
  12. 8:19 · watch on youtube.com↗

    What I'm trying to say is Anthropic is not above doing the same thing. They could cut off access to Cursor as well.
  13. 3 min later
  14. 11:12 · watch on youtube.com↗

    they already kind of used all the data on the web violating copyright laws left and right, not really caring about that, but now when other companies are trying to distill their models and use them for training data, now it's a problem
  15. 1 min later
  16. 12:06 · watch on youtube.com↗

    Well, they're definitely not going to slow down if it's going to hurt their own company.
  17. 1 min later
  18. 13:14 · watch on youtube.com↗

    Even if you're just going based on the API pricing, you can't just go one-to-one because not only is there API pricing included, but if a model is the same price as another model, but it's just more efficient, token efficient, then you're going to be paying less for it anyway.
  19. 13:38 · watch on youtube.com↗

    And that's part of the reason why companies might prefer open weight models, even if they come from China, because you can just host the models yourself.
  20. 1 min later
  21. 14:36 · watch on youtube.com↗

    I'm certainly sensing a lot of desperation from OpenAI and Anthropic, and I don't think SpaceX or xAI are the saints here either. But at the very least, I am glad that there is a lot of competition going on right now.
  22. 14:53 · watch on youtube.com↗

    And while I do think that this was the wrong move from OpenAI to do this, because it's going to just push people more towards Chinese models, and also just lose a lot of trust in American companies
  23. 15:19 · watch on youtube.com↗

    it's hard for me to imagine a future 5 to 10 years from now where open weight models aren't getting more and more usage.
  24. 15:27 · watch on youtube.com↗

    At the very least, it's getting harder and harder to trust the labs, if that was even possible in the first place.