I'm working on a similar simulation game, and considered possibly using an LLM for AI, but packaging a 300GB model into the game isn't practical. Alternatively, I wonder what kind of latency would be experienced in deferring to an OpenAI API, for the intelligence? What kind of bottleneck would occur if you're constantly querying an API...possibly every few game loop iterations? I'll probably stick to home-grown AI behavior for now, but it's an interesting concept for people with easy access to large amounts of compute.
Love these paper walkthrough videos. Really appreciate your overview and expertise.
This is so cool I’ve been trying to do something like this. Now I have a reference to look back at.
I'm working on a similar simulation game, and considered possibly using an LLM for AI, but packaging a 300GB model into the game isn't practical. Alternatively, I wonder what kind of latency would be experienced in deferring to an OpenAI API, for the intelligence? What kind of bottleneck would occur if you're constantly querying an API...possibly every few game loop iterations? I'll probably stick to home-grown AI behavior for now, but it's an interesting concept for people with easy access to large amounts of compute.
Wow... Very interesting, makes one wonder if they are somewhat sentient
This is the evidence what we are living in the matrix😂
As above so below
This is more just like a more complex dwarf fortress