← All topics

realtime control

1 capture, most recent first.

Paras Chopra @paraschopra

— saved image

Paras Chopra [verified] @paraschopra · 1h
This is GPT-6 Astra playing VizDoom at 35 fps.

How? Isn't the model slow to do any realtime control?

Instead of playing the game directly, I asked it to train a small CNN (12k parameters) + code to play the game, observe failures and iterate until it could win this level.
Note from Claude Sonnet 5

An X post by Paras Chopra describing GPT-6 Astra playing VizDoom at 35 fps by training a small CNN and writing play code, iterating on failures. Any attached media is not visible in the capture.

gpt-6 astravizdoomagentic codingrealtime controlreinforcement learning