Yeah, that would be swell

God I wish

Those are starting to become a thing

clawbox.com/ai-assistant-box

futureproofhomes.net/products/nexus

I would use this, but how can one trust that there is any privacy on a product from any of the big tech companies?

I think we just (rightfully) lost faith in tech products overall by now.

I would say no, because even physical devices you purchase have to come with subscription plans or in some way talk back to the manufacturers servers that isn’t in your interest.

The pitch here was that it wouldn’t have to speak to the manufacturer’s servers. The device, “AI” and data are yours. I may have not been clear here. As someone has already said in these comments, there are models, that exist, designed to run on phones and home servers, they just dont have a ton of comoute power and are limited in that way but i think that’s kinda the point. Most people arent going to be writing code or making heavy requests. Most people would use it to help them with everyday things

The things openai is making comerciales about, like helping figure out your commute and daily needs. Maybe even tracking purchases for tax purposes, some home automation,

Theres is a doctor running his own LLM on a single server cross checking FDA approved drugs and potential uses in un treated diseases. Its not fast but working steadily at its purpose. Maybe your usecase is closer to this.

Im just saying theres no question the current state of “AI” is a terrible consolidation of the worst parts of capitalism and shouldn’t be allowed to continue, but how could LLMs be used responsibly and usefully?

You can run some pretty neat local AI agents on consumer grade hardware without much issue. There are also new models that will run on phone hardware. They won’t write complicated programs, but they work well for assistant things.

Ew, no

Ok but why ew?

Probably less so.

The hardware to run it locally would be fairly expensive and would require using a very simple model compared to alternatives. Also much more wasteful if you were only using that hardware for AI use, as you’d be distributing the hardware out rather than centralizing so it’d be often idle and when replaced create more waste than a centralized server rack.

Also, additional per user post-training seems to me both wasteful and not necessary. In context learning is often much more powerful but frequently overlooked.