korrents

korrents · video

How I use LLMs

Andrej Karpathy · 2h 11m · youtube.com ↗

17 korrents from this recording

1h2h
Andrej Karpathy 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. 0:08:19 · watch on youtube.com↗

    the zip file is not exact the zip file is lossy and probabilistic zip file
  2. 2 min later
  3. 0:10:12 · watch on youtube.com↗

    roughly speaking you should think of these uh models is kind of like a little bit out of date because pre- training is way too expensive and happens infrequently
  4. 6 min later
  5. 0:16:32 · watch on youtube.com↗

    anytime you are switching topic I encourage you to always start a new chat
  6. 1 min later
  7. 0:17:36 · watch on youtube.com↗

    think of the tokens in the context window as a precious resource um think of that as the working memory of the model and don't overload it with irrelevant information and keep it as short as you can
  8. 3 min later
  9. 0:20:29 · watch on youtube.com↗

    if the intelligence is not good enough for you and you're using this professionally you may really want to consider paying for the top tier models that are available from these companies
  10. 22 min later
  11. 0:42:10 · watch on youtube.com↗

    I think it's incredibly cool and really interesting and kind of went under the radar for a lot of people even though I think it shouldn't have
  12. 6 min later
  13. 0:48:38 · watch on youtube.com↗

    treat this as your first draft treat this as papers to look at um but don't take this as uh definitely true
  14. 10 min later
  15. 0:58:09 · watch on youtube.com↗

    I feel like llms make a lot of reading very dramatically more accessible than it used to be before
  16. 1 min later
  17. 0:58:52 · watch on youtube.com↗

    this is extremely helpful I encourage you to experiment with it and uh don't read books alone
  18. 8 min later
  19. 1:06:27 · watch on youtube.com↗

    I would be hesitant to potentially recommend the use of these tools uh if people aren't able to like read it and verify it a little bit for themselves
  20. 21 min later
  21. 1:27:22 · watch on youtube.com↗

    the big takeaway here is don't type stuff out use voice it works quite well
  22. 18 min later
  23. 1:44:58 · watch on youtube.com↗

    you probably want to talk to an actual doctor as well but I think this is a really good first draft and something that maybe gives you things to talk about with your doctor
  24. 1 min later
  25. 1:46:15 · watch on youtube.com↗

    it's really upsetting to me that companies put all this stuff in your um in your food or cosmetics and stuff like that when it really doesn't need to be there
  26. 14 min later
  27. 2:00:31 · watch on youtube.com↗

    I find that this always increases the accuracy of LMS so kind of that's a I think a general good strategy
  28. 3 min later
  29. 2:03:03 · watch on youtube.com↗

    if you're learning different language I would not use a different translator other than Chachi PT it understands a ton of nuance it understands slang it's extremely good
  30. 5 min later
  31. 2:07:42 · watch on youtube.com↗

    I think Chachi by far is a very good default
  32. 1 min later
  33. 2:08:45 · watch on youtube.com↗

    try without reasoning first and if your model is not solving that kind of kind of a problem try to switch to a reasoning model