AI on a Pentium 2 PC

[MrBcx]

I'm sure the team had fun tinkering but I'd bet that their retro
project is unlikely to yield any meaningful new discoveries.

A 1B Llama LLM is a weak model and getting the inferencer to "run"
it, at 1 token per second, is pain on top of pain on top of pain.

[jbk]

"Someone Experimented With a 1997 Processor and Showed That Only 128 MB of Ram Were Needed to Run a Modern AI" https://indiandefencereview.com/someone-experimented-with-a-1997-processor-and-showed-that-only-128-mb-of-ram-were-needed-to-run-a-modern-ai/

this should be an interesting read for MrBcx