gemini-4POST
Google uses Gemini 4 Argon to generate robot training simulations from human action videos
Google AI shared an experiment using Gemini 4 Argon to generate training simulations from human action videos, showing a robot learning a beanbag-toss task, with the author urging questions about where the simulation's physics parameters come from and when they fail.
飞叔
@RyanJ_KOLArchived 21 days agoOriginally posted 2026.10.14
飞叔 @RyanJ_KOLOriginally posted 2026.10.14
Google AI 今天分享了一项实验：用 Gemini 4 Argon 从人类动作视频生成训练模拟，展示了机器人学习投豆袋游戏的流程。 我更想看模拟里的“看不见的假设”：豆袋重量、板面摩擦、碰撞后的弹性，这些参数分别来自测量还是默认设置？ 验收时，可以让模型列出参数来源，给尚未实测的值标上范围，再换几组条件，记录动作在哪些情况下失效。 自动生成模拟之后，值得投入的下一步，是把模型猜出的环境条件变成能测量、能调整的项目。这样才知道该去现场补哪份数据。 #AI #机器人 https://x.com/GoogleAI/status/2110083927620591769
This post quotes another post: “Google demo shows Gemini 4 Argon speeding up robotics research”
x.com/ryanj_kol/status/2110161107617006010Archived 2026.10.14 10:24