This will probably not surprise you to learn that generating AI was a subject with white in the world of robotics. There are a number of Different ideas floating around On the best ways to embrace emerging technologies, from natural language controls to design. I ask the question of generative AI to Deepu Talla, vice-president and managing director of Nvidia de Embedded & Edge Computing, during a recent visit At the headquarters of South Bay.
“I think it speaks in the results. You can already see the improvement in productivity, ”said the executive to me. “He can compose an email for me. It’s not really good, but I don’t have zero. It gives me 70%. There are obvious things that you can already see who is certainly a better step function than the way things were before. Summarizing something is not perfect. I’m not going to let him read and sum up for me. So you can already see certain signs of productivity improvement. »»
It turns out that Nvidia was only a few weeks old about her news concerning the subject. Roscon’s announcement comes next to several other bits of news linked to its various robotics offers, including the general availability of Nvidia Isaac Ros 2.0 and Nvidia Isaac SIM 2023 platforms.
The systems adopt a generative AI, which should go to the acceleration of its adoption among the Roboticians. After all, as Nvidia notes, some 1.2 million developers interfaced with the Nvidia AI and Jetson platforms. This includes some big customers such as AWS, Cisco and John Deere.
One of the most interesting bits here is the AI Lab generative jetson, which gives developers access to models of large open source. The company writes:
The NVIDIA Jetson Generative AI LAB offers developers access to optimized tools and tutorials for the deployment of open-source LLMS, diffusion models to generate breathtaking images in an interactive way, vision language models (VLM) and vision processors (VITS) which combine the vision of AI and natural language treatment to provide a complete understanding of scene.
The arrival of these types of models can follow a way to help systems determine a line of conduct in circumstances on which they were not already formed (only, the simulation takes place so far). After all, while spots like warehouses and factory floors are more structured than, say, a highway, there are still countless variables to face. The idea is both to be able to adapt to the volley and to offer a more natural language interface for systems.
“The generative AI will considerably accelerate the deployments of AI at the cutting edge with better generalization, ease of use and higher precision than possible,” said Talla in a statement linked to today’s news. “This greater software expansion of our Metropolis and ISAAC frames on Jetson, combined with the power of the models of transformers and the generative AI, responds to this need.”
The latest versions of platforms also make improvements to perception and simulation.