This year’s COMPUTEX forum “AI In Action” began on Wednesday morning, bringing industry experts from NVIDIA to Google to discuss “Robotics & Edge AI Application.”
From companion pets to everyday assistants, factory production lines to hospitals, today’s forum proved that Edge AI has already ushered in the future of robotics.
First up, Deepu Talla, NVIDIA’s Vice President of Robotics and Edge AI, presented “Physical AI and Robotics: The Next Multi-Trillion AI Opportunity.”
After over a decade in the industry, Talla said, “This is the first year I feel the future is coming to the present.”
Talla highlighted breakthroughs in harnessing digital AI from large-language models such as ChatGPT for application in physical AI, notably for manufacturing and production.
NVIDIA acts as a platform company, assisting 350 Taiwanese companies in robotics, across hardware, software, simulation and system integration.
Using NVIDIA’s digital twin technology to test robotics safely, quickly and cost-effectively, physical AI is now being deployed from Kaohsiung’s smart city design by Linker Vision to Pegatron teaching employees complex tasks using AI agents.
“The age of industrial AI is here, pioneered by the industry leaders of Taiwan, powered by [NVIDIA’s] Omniverse,” Talla said.
Next, Ed H. Chi, Vice President of Research at Google DeepMind, discussed the “Google Gemini Era: Bringing AI to Universal Assistant and the Real World.”
Google Search’s AI overview sees over 1.5 billion monthly users, bringing generative AI to more people than any product worldwide, Chi said.
Chi outlined three advancements in Gemini robotics: generalizing, adapting to new scenarios using world-understanding capability without retraining; interactivity: eliminating data reliance to respond quickly to changing commands and conditions; and dexterity: performing complex tasks such as tying shoelaces.
The key to Gemini is perception, planning, reasoning and real-time execution of natural commands: “this is the first time in history this integration has been achieved in a single model,” he said.
Giovanni Campanella, a Sector General Manager at Texas Instruments, then took the stage for “The future of humanoid robotics with advanced semiconductor technology.”
Texas Instruments believes “safety is the foundation of every robot design,” integrated from sensors to semiconductors, ensuring robotics can work safely alongside humans.
Ongoing innovation now focuses on creating power-efficient AGI, real-time communication interfaces and smooth control for precise motion.
“Humanoid is the present,” he said; “we are already seeing prototypes of humanoids that are walking, working side-by-side with humans, performing complex tasks.”
Lastly, Miller Chang, President of Embedded Sector at Advantech presented “Edge AI-powered Robotics in Action.”
Robotics and Edge-AI were previously “two separate things,” but as isolated Edge computing evolved into a complex ecosystem of Edge AI combined with IoT, Edge AI now forms the “core building block” for autonomous systems, he said. Partnering with NVIDIA, Intel and Qualcomm, Advantech developed the Robotic Suite, an all-in-one corporate package.
Finally, Chang let the robots do the talking, bringing two pet robots on-stage that moved with the speed and agility of a dog, performed tricks and followed commands in both English and Mandarin.
