388
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 06 Sep 2026
388 points (95.8% liked)
Technology
87935 readers
3414 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 3 years ago
MODERATORS
This is called burying the lede.
If Microsoft knows this cloud-based AI thing is a bust and people will be moving to local models that are not controlled by Microsoft or OpenAI or Anthropic, then it's only a few months before Satya tells his investors buddies and the bubble pops.
I find this unlikely... IMO I think they are hedging in both directions.
Personally I like the idea of local AI. open source models that you an control, decide, and use smartly, sort your things etc... and not pay an absurd cost or link up.
Personally I think MS is actually betting on the opposite. IE they had a prototype windows that was basically just an AI prompt, no start menu etc... and applications just "streamed".
IMO I think that's more of the goal, use the data center buildout to drive up hardware prices, until people cannot afford a basic PC. Sell thin clients... then sell us "access" to compute in a form where we own nothing, but pay by the processes used etc...
I'm pretty sure local models still induce pyschosis and hallucinate. Maybe the cat's out of the bag, but they still suck.
They absolutely do suck, but I can see some practical uses for them. IE I can see LLM's as useful in localized specialized uses. for me I'd really like basically LLM as a like calendar, scheduling kind of thing, also say translating home assistant commands through a voice assistant to make it figure out actual intent etc..., nothing super mission critical, but also mostly things I wouldn't want any 3rd party service that could change the rules for me, share them to others etc...
I agree it's a hedge, but I think it's a pretty big admission that the writing is on the wall.
Given how much Windows has invested in cloud AI, that they are encouraging developers--the sweetest plum, the people most likely to pay for cloud AI--a local environment is a pretty big admission. This is the sort of thing corporations that have the ability to control markets don't do unless they think the market is going to leave them behind.
64GB would be on the short side for local AI stuff today, 128 with some 80-96 shared for vram is probably a better start. IMO.