Jim Fan, Senior Research Manager and Lead of the Embodied AI division Generalist Embodied Agent Research (GEAR) Lab, Nvidia, underlined how artificial intelligence (AI) robots in the future would be educated and operate. According to the CEO, shortly embodied artificial intelligence agents will be formed in simulation where they will study and acquire knowledge in certain activities. He also underlined how soon whole towns, homes, and factories would be sent to simulations, thereby enabling training AI on real-world situations. Future artificial intelligence robots might possibly have a hive consciousness, he further hypothesised.
Jim Fan Shares by Nvidia on Future AI Robot Training
Fan noticed in a post on X (previously known as Twitter) that the City of Tokyo has made available for download a 3D digital twin of the whole city in a high-resolution point cloud. According to the Nvidia executive sharing the information, this trend of bringing real-world organizations to digital simulations is just going to becoming more prevalent in the future.
“More and more cities, homes, and businesses will inevitably be modeled into simulations,” Fan added.
Drawing on the example, Fan also pointed out that robots won’t be taught in solitude going forward. Robots are taught in controlled environments right now where they may pick and drop stuff, learn how to move about, recognize objects, and do jobs. Many have argued, nevertheless, that this kind of instruction does not equip the robots for the several obstacles they would encounter in the actual world.
Presenting a fix for the issue, the Nvidia research manager said that shortly embodied AI bots will be modeled as an iron fleet. Especially embodied artificial intelligence agents are AI systems connected with a real body or a virtual representation that allow them to see the environment like humans or animals.
According to Fan, future artificial intelligence robots would be scaled over a vast cluster and housed in real-time graphics engines. Such approaches will also generate trillions of excellent training tokens, he said. Extending the hypothetical vision, he remarked, “The majority of embodied agents will be born in sim, and transferred zero-shot to our real world when they are ready.”
Although this would seem to be science fiction, Fan, head of the embodied AI team at Nvidia, is sure that this will be the path to equip very powerful robots in not too distant future. Nvidia seems to be already headed in that route as well. He said that before materializing in atoms, the company’s Santa Clara headquarters buildings were “designed and rendered in Omniverse, a GPU-accelerated graphics platform.”
