The ultimate goal of artificial intelligence giants: Physical Artificial Intelligence

Technology and artificial intelligence giants are preparing to implement physical artificial intelligence step by step. In recent weeks, Nvidia has released physical artificial intelligence focused world model family While introducing, Google To establish a new team for artificial intelligence models that can simulate the physical world focused. Also OpenAI He started to establish his robotics team. he explained.

First, let’s talk about OpenAI’s announcement. According to the information shared, OpenAI’s robotics team aims to combine the company’s advanced artificial intelligence models with physical robotic platforms. Driven by the vision of developing robots that are “general purpose” and can work in “dynamic, real-world environments”, OpenAI plans to bring these projects to full-scale production in the future.

OpenAI Hardware Director Caitlin KalinowskiIn a post on the social media platform X, he stated that the company has started to establish a special team that will develop robotic sensor systems and mechanical components. In the new job postings, positions such as “mechanical engineers responsible for the design of components such as motors, gears and actuators”, “electrical engineers who will develop sensor systems for robots” and “project managers who will focus on testing robot prototypes” attract attention.

The announcement that Google will establish a new team for artificial intelligence models that can simulate the physical world came from Tim Brooks, who joined Google DeepMind in October and was previously the co-president of OpenAI’s artificial intelligence video tool Sora.

shared by Brooks according to job postingsthe team to be formed is planned to work together with Google’s Gemini, Veo and Genie teams. The team will work on these projects to simulate the physical world, solving critical new problems and taking the models to the highest level of computing.

World models will power a wide range of areas, including visual reasoning and simulation, planning for embodied agents, and real-time interactive entertainment, the company said. According to the statement, Brooks’ new team will work to develop “real-time interactive production” tools on top of the models they have created and will explore how to integrate their models with existing multi-modal models such as Gemini.

We can say that Nvidia is one step ahead of Google and OpenAI when it comes to world models. Nvidia, Cosmos World Foundation Models It has made the world models family called openly available for use. announced. This family of world models can predict and produce physics-aware videos. Built specifically for physical AI research and development, Cosmos WFMs can create physics-based videos from a combination of inputs such as text, images and video, as well as robot sensor or motion data.

According to Nvidia’s statement; Companies such as Waabi, Wayve, Fortellix and Uber have stated that they will pilot Cosmos WFMs for various use cases. These companies can leverage Cosmos WFMs for purposes such as video search, video curation, and building AI models for self-driving vehicles. Let us also remind you that Nvidia has artificial intelligence and simulation tools that support robot learning and humanoid robots.

Increasingly popular among AI companies, world models can be used to create interactive media, replacing games and movies. Likewise, these models can also be preferred to run realistic simulations such as training environments for robots. We have seen many initiatives taken to create a world model so far. last month Receiving an investment of 32 million dollars Decart and in July Receiving an investment of 100 million dollars World Labs is the first thing that comes to mind.