zlacker

He has speculated a few times about an LLM based OS. That’d be way cool.

>>next_x+(OP)
Honestly? Sounds like a nightmare. I mean, some LLM integrated into a OS, ok, might make sense, but the OS based on LLM is not something I would want with the current state of the art.

>>next_x+(OP)
Why?

>>jjtheb+R
Non-deterministic outputs... For one (rather important) thing.

>>llamaL+U1
I can think of lots of reasons why non-deterministic outputs at the OS level is a bad idea, but what are the benefits?

>>lolind+k2
Not to jump in for someone else but your use of “OS level” prompts me to opine: I think the features of a meaningful new OS would extend far beyond the programmatic level of the kernel, the drivers, the dependencies, etc. A “new OS” could just be Linux with some cool UX innovations on top enabled by ensembles of lightweight, purpose built LLMs. Think window management, file management, password management, etc.

For one potentially compelling example that happily (sadly?) isn’t using LLMs: the SimulaVR people are developing their own Linux fork of some kind, claiming it’s necessary for comfortable VR use for office work. And I sorta believe them!

>>lukan+y
Features like function calling are moving in that direction. Microsoft also seems to have plans to deeply integrate LLMs into its OS and if they do a good job it could become a primary way to interact with its features and programs. Considering the progress made on image generation models I could image a special purpose model that is specifically trained on operating APIs and producing good results. The big hurdle would be building the APIs that don't exist for the tools that people like to use. I'm sure there are interesting ways you could think of generating labeled data for actions in various programs.

>>next_x+(OP)
> LLM based OS

Isn't that is what Rabbit R1 is? https://www.rabbit.tech/

>>llamaL+U1
non-deterministic outputs from what? I'm misunderstanding your idea.

>>jjtheb+RL1
My point was that non-deterministic outputs make LLM's fundamentally a BAD foundation for something like an OS.

Natural language interfaces belong at the periphery, as the interface between the human and the machine. Other than that, I want my computers dumb as rocks, really fast, any totally predictable - which is basically the opposite of what you get from LLM's.