Aether AI

Aether AI is the company founded by Huang Biwei and discussed in 哪条路线，才能通往「世界模型」的终局？｜对话黄碧薇：Aether AI 创始人. The source presents the company as building Causal World Models, with a first landing area in Embodied AI as a generalized robot brain.

Strategy

Aether AI’s route is to combine causal research, large-scale data, simulation, video-model learning, and robotics deployment rather than relying on imitation learning alone. Huang says the company uses simulated data, egocentric data, video data, and teleoperation data, with teleoperation serving as the last-mile mapping from physical regularities to robot embodiment.

Operating Notes

• First-version training is described as needing roughly seven to eight thousand hours of data and several hundred GPUs.
• The team reportedly had about four hundred cards at the time of the episode.
• Financing is described as roughly 20 million USD, mainly for compute, data, and hiring.
• Hiring needs include causal algorithm talent, video-model training talent, and robotics full-stack talent.

Connections

• Huang Biwei — founder and source guest.
• Causal World Models and Causal AI — core technical direction.
• World Models, Embodied AI, Vision Language Action Models, and World Action Models — surrounding robot-modeling landscape.
• Frontier Model Scaling — compute and data planning context.