独家｜北大董豪：「仅停留在数据层面的Scaling Law，教不出通用机器人」

Instead, a two-dimensional Scaling of 'data volume × task volume' is essential for achieving embodied AGI.

Key Points

• Current model technology iterations in embodied intelligence are slowing down.
• There is ongoing debate regarding the direction of these model developments.
• Dong Hao emphasizes the need for a broader scaling approach for effective AGI.