Fei-Fei Li, a researcher and founder of World Labs, believes that the future success of artificial intelligence depends on the development of spatial reasoning, not just language processing. She is convinced that machines must learn to understand movement, distances, and physical relationships in order to become partners for humans.
According to Li, current language models are inefficient as they cannot operate in the physical world, determine the locations of objects, and predict the consequences of their movement. Therefore, creating internal models that correspond to physical laws is key to the development of artificial intelligence.
Creating World Models
Fei-Fei Li proposed the idea of creating world models that will be generative, multimodal, and interactive. Her team at World Labs is developing a prototype system called Marble, which can create three-dimensional scenes based on multimodal queries.
In the future, according to the scientist, artificial intelligence capable of reasoning from spatial data may find applications in robotics, complex tasks, and scientific research, contributing to advances in machine thinking.
At the National Museum of Cardiff (UK), an artist under the pseudonym Elias Merrow installed a painting created by artificial intelligence, which for some time did not attract the attention of staff.
Fei-Fei Li emphasizes the importance of developing spatial reasoning for artificial intelligence to make it more flexible and interactive with the physical world. Her Marble project could serve as an excellent example of this approach, opening up new opportunities for the application of artificial intelligence across various fields.