Google DeepMind's Gemini 4 Argon learns tasks from video and transfers them to Apptronik's Apollo

Google DeepMind showed its new Gemini 4 Argon model watching a video of a person playing cornhole, building a simulation of the task and passing the learned motion to Apptronik's Apollo humanoid, which then landed seven throws in a row. Google says the video-to-simulation pipeline can shrink weeks of manual setup to a few hours.

Summary written by DroidBazaar. The original article belongs to its publisher.