A mechanical hand showcased at the Robot Mall in Beijing represents the future of AI integration in robotics. In a significant move towards enhancing artificial intelligence, Alibaba Cloud is investing in a new type of AI that aims to more effectively mimic real-world scenarios, diverging from the traditional text-based approaches employed by systems like ChatGPT.
This approach focuses on developing “world models” that utilize multimodal data, including videos and real-life physical interactions, rather than relying solely on large language models. To support this initiative, Alibaba has committed 2 billion yuan (approximately $290 million) to ShengShu, a startup specializing in AI video generation through its tool, Vidu. This funding round also saw contributions from TAL Education and Baidu Ventures.
ShengShu has previously secured additional investments, including a 600 million yuan round from Qiming Venture Partners. The startup’s focus is to create a “general world model” that can seamlessly integrate digital environments, such as video games, with the physical world relevant to robotics and autonomous driving. Founder Zhu Jun noted that the goal is to enhance AI’s ability to model and predict real-world behaviors by connecting perception and action.
In the competitive landscape, Alibaba has recently invested in other startups like Tripo AI and PixVerse, which are also working on AI that emphasizes physical interaction and video generation. As the industry evolves, experts highlight the necessity of world models in robotics to facilitate the development of intuitive AI systems that mimic human intelligence effectively.
Why this story matters:
- Highlights the shift from text-based AI models to more integrated world models in robotics.
Key takeaway:
- The combination of multimodal data and world models is crucial for advancing AI’s ability to replicate real-world interactions.
Opposing viewpoint:
- Some industry experts argue that reliance on world models alone may not be sufficient to achieve true human-like intelligence in AI.
