Yandex has begun developing Russia's first Physical Artificial Intelligence (Physical AI), which will be able to deeply understand the material world and interact with it. This was reported by the Yandex press service.
The experience accumulated on roads and indoors, combined with Yandex technologies, will allow robots and autonomous vehicles to comprehensively process multimodal data: image, video, sound, text. This will bring their perception closer to human perception.
As part of the project, a VLA (Vision-Language-Action model) has already been created and trained, which transforms voice and text commands, as well as data from cameras, into robot actions. More than 10 basic actions are supported, such as "take", "put", "transfer", and in the future there will be hundreds more. Also, Yandex Robotics is developing Yandex RMS, which allows robots to choose the optimal combinations of actions to perform tasks, and in case of data shortage — to request them from related systems.