Local LLMs are the future, and one of the reasons I think the data centre furore is just going to end in a market crash.
LLMs that can reliably be used for control problems are the future, and I think classic/deep RL has generally been overlooked for years for a whole host of problems by wider industry because it felt inaccessible. The first thing I thought of when I saw Jev (and then Laya), was "this might move the needle in a really, really interesting way".
Local LLMs that can reliably be used for control problems smash through a lot of barriers I'm interested in, and this intrigues me a lot. Guess I'm about to become a big Laya fan if it can run on this kind of hardware to this performance.
Great job! Did you finetune your own Laya for the snake game or what?
This looks like a local AI model playing Snake -- is that correct? The article offers no explanation.
The linked github project [1] contains more information.
[dead]
that's not a local LLM. If it's local, it doesn't matter in this case. Laya is a System 1 "AI", namely works like a classifier, given a state and questions, it shoots probabilities for each. I publish an episode tomorrow about Laya and Typesafe AI on https://www.youtube.com/@DataScienceatHome
Stay tuned ;)