Xiaomi open-sources AI model for autonomous driving and robotics
What's the story
Xiaomi has open-sourced a groundbreaking foundation model called MiMo-Embodied. The move comes as part of the Chinese tech giant's efforts to bolster its artificial intelligence (AI) capabilities. MiMo-Embodied is touted as the "first open-source vision-language model" that seamlessly integrates tasks related to autonomous driving and embodied AI. Technical reports and project files can be found on platforms like Hugging Face and GitHub.
Model capabilities
MiMo-Embodied's performance in AI tasks
Xiaomi claims that the cross-embodied model, MiMo-Embodied, has achieved state-of-the-art performance in both autonomous driving and embodied AI tasks. The model showed "superior" results in task planning, affordance prediction, and spatial understanding for embodied AI. For autonomous driving, it excelled at environmental perception, status prediction, and drive planning. This dual capability makes MiMo-Embodied a unique addition to the field of artificial intelligence models.
Tech trend
Rise of embodied intelligence in China
The rise of embodied intelligence, which integrates AI into physical systems, has gained traction with the rapid development of humanoid robotics in China. Xiaomi has been at the forefront of this trend, having unveiled prototypes of a 4-legged robot dog and a humanoid robot in 2021 and 2022 respectively. The company has also invested heavily in these areas to further its technological advancements.
Talent acquisition
Xiaomi's recent recruitment
Xiaomi's announcement of the open-source model comes after its recent recruitment of Luo Fuli, a former core researcher at Chinese AI start-up DeepSeek. She joined Xiaomi several months after the company reportedly approached her.