• 0 Posts
  • 12 Comments
Joined 3 years ago
cake
Cake day: June 15th, 2023

help-circle


  • yea, i’m surprised, 32GB is goddamn ridiculous for anything, let alone for a shitty hp branded autocorrect

    32GB is actually considered the bare minimum for most of the common locally run LLM models. Most folks don’t run a locally run LLM. They use a cloud service, so they don’t need a huge pile of RAM locally. However, more privacy focused or heavy users with cost concerns might choose to run an LLM locally so they’re not paying per token. With regards to locally run LLMs, this would be comparable to renting car when you need it vs buying one outright. If you only need a car once a year, renting is clearly the better choice. If you’re driving to work everyday then clearly buying the car yourself is a better deal overall.

    You are perfectly fine not liking AI, but you’re also out-of-touch if you think 32GB is too big for anything. Lots of other use cases need 32GB or more and have nothing to do with AI.

    I agree with your frustration with subscription laptops. I hope people don’t use it.





  • Sure, but the powers in Tallinn have to make this statement first to head off a diplomatic crisis. If Estonia had just started shooting Russians it would have been a much harder position diplomatically. There’s no room for Russia to claim “we accidentally stepped on your territory” because Estonia has now made the statement publicly what the consequences for Russian soldiers is that do that. Any Russian soldiers that do that (or their commanders) now have a full understanding they’ll be shot.

    When the first shooting actually occurs, Russia can’t present any kind of credible complaint about “you shot our soldiers that were just lost!”. That would show incompetence on the part of Russian commanders, which is something Russian’s are even more loathed to admit.