Comment by MatthiasPortzel
13 hours ago
People have done experiments like this before.
Minecraft isn't a intelligence problem. An LLM is more than smart enough to play Minecraft. It's a robotic problem. Even a normal human player can click several times per second, can press two different buttons at the same time, etc. Mineflayer gives you an API that is more limited than a human in some ways (can't see far away from you) and less limited in others (can see through blocks) (if I remember correctly, I could be thinking of Baritone or a different project). Robotics or playing a video game requires a tight continuous feedback loop. When there's 500ms of latency in order to ask the model for every mouse movement, it's no wonder fighting a spider is scary.
Fun fact, Mineflayer, the Minecraft bot interaction API used by YCombinator for these tests, was originally developed by Andrew Kelley (before he started Zig).
Minecraft isn't a problem at all, it's a sandbox game where it's implied you make your own fun. Depending on the levers you give the agents, it may or may not be a robotics problem, but it most of all depends on the task you give the agents. The task they were given in this case was pretty lame in my opinion ("build a civilization"), and it's no surprise they were bureaucratic and ineffective. It pretty much goes with the genre.
I'd imagine LLM could trivially write code that fights at least single opponent competently, and then just run it ?
It wrote scripts for fishing, should work just fine with combat
The human body has the ability to automate some muscular movements so the brain is out of the loop. Maybe these LLMs should be allowed to write in a small DSL that can trigger events out-of-band (something like "Oh fuck there's a spider, I'll set up a click spam loop and check in every 500ms").
I think they're doing exactly what you propose? Unless I'm misunderstanding what the Python "fishing scripts" with the timings and timeouts are doing.
https://labs.ycombinator.com/projects/games/index.html#fishi...
Mineflayer does not strictly limit view distance, you just have to change the setting for it from the default value.
Get the LLMs playing GT:NH.
Use traditional game AI for the micro, LLMs for the macro