Jev 驱动的实时 Doom 演示
发布线程展示游戏场景中的实时决策效果。Diogo 说明演示约每秒调用 10 次 Jev,并附有 2 分 43 秒视频。

作者提出将大语言模型视为大脑、Jev 视为虚拟身体反射的设想,并尝试用 Jev 为 3D 角色选择情绪反应,通过面部和动作表达细微情绪变化。
If large language models are the brain, could Jev be the reflexes running throughout a virtual body? This started with a simple idea: using Jev to choose emotional reactions for a 3D character. I wanted it to express subtle changes in emotion through its face and movements. That