gpt-6VIDEO
用 GPT-6 Astra 训练小 CNN 打 VizDoom，跑到 35 fps
作者没让 Astra 直接实时操控游戏，而是让它写代码训练一个 1.2 万参数的小 CNN，观察失败并迭代，最终以 35 fps 通关该关卡。
Paras Chopra
@paraschopra2026.09.29收录原推发布于 2026.09.28
Paras Chopra @paraschopra原推发布于 2026.09.28
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.
x.com/paraschopra/status/2104414175049269633采集于 2026.09.29 10:16