Founder and CEO of NVIDIA Jensen Huang delivered the first Keynote address of COMPUTEX 2024 on Sunday night, in advance of the official exhibition opening on Tuesday 4th June, with a charismatic presentation that illuminated how generative AI has come to impact almost every industry in the world today and unveiled NVIDIA’s plans to drive the technology to new heights in the future.
The arrival of generative AI has heralded a “new Industrial Revolution” Huang told a packed hall at National Taiwan University Sports Center as he showcased the company’s latest Blackwell GPU architecture. Capable of performing at up to 20,000 FLOPS, the new generation of AI chips represents a 1,000-fold increase in computation speed over the past eight years, vastly outperforming Moore’s law over the same time period.
The Blackwell chips are now in production and powering “the most complex, highest performance computers the world has ever made,” said Huang.
The advent of accelerated computing, in which GPU chips utilize parallel processing to drastically increase computational speed and cost efficiency, has surmounted the obstacles presented by stagnating CPU performance scaling, and has driven the shift from perceptive AI models capable of pattern recognition to generative models that produce new data and content.
Faster performance means greater cost efficiency, explained Huang, which then drives uptake by developers and manufacturers. The growing market and install base then allows NVIDIA to increase and upscale their R&D, leading to even faster speeds and achieving a virtuous cycle.
To complement their next-generation hardware offering, the company is introducing NIMs, or NVIDIA Inference Microservices, that allow developers to directly implement pre-trained AI models into their applications across the cloud, data centers and workstations. Functioning as an “AI-in-a-box”, NIMs provide an accessible and standardized way for developers to integrate generative AI models including text, image, video, and speech into their work, boosting their productivity.
Huang concluded by detailing a blueprint for the future. “The next wave of AI is physical AI”, he said, as robotic technologies come to understand, interact with, and shape the physical world alongside humans. This is already happening today, as NVIDIA’s technology is being used by companies such as Foxconn to design, model, and operate real-world factories and warehouses.
Self-driving cars and humanoid robots are two of the high-volume products that NVIDIA are focussing on to power the coming wave of physical AI, with partnerships with Mercedes and Jaguar Land Rover launching in the next two years. Meanwhile, AI is expected to lead a paradigm shift in the field of humanoid robots in the near future, because their rapidly advancing cognitive and physical capabilities will be perfectly adapted for the world which humans have already built around themselves.
