Nvidia has been a frontrunner in the world of AI. The tech giant’s GPUs have been used in developing tools like OpenAI’s ChatGPT. And recently, the company unveiled new AI chips that will add to its dominance in the AI space. During the event, Nvidia CEO Jensen Huang said that the future will be all about robots, and we will get to see more of them in the coming years. This isn’t the first time that the Nvidia CEO has talked about the promising future of AI and robotics. Recently, he said that in the next five years, AI could excel in various tests presented by the computer science industry.
Nvidia unveils new AI chips
According to a Bloomberg report, Nvidia introduced a new generation of its AI chips called Blackwell. The Blackwell processor design was unveiled during the GTC conference in San Jose, California. The chips promise significant improvements in processing speed for both training and inference tasks. Equipped with 208 billion transistors, the Blackwell chips are slated to power upcoming computers and products for major data centre operators like Amazon, Microsoft, Google, and Oracle.
Additionally, The initial Blackwell chip, known as the GB200, is set to be released later this year. Nvidia also encouraged its customers to upgrade by offering more potent chips. The report also mentions that many businesses and software developers are still struggling to obtain the current “Hopper” H100s and similar chips. However, during the conference, Nvidia CEO Jensen Huang expressed the need for larger GPUs and said, “Hopper is great, but we require bigger ones.”
The report also revealed that the new chip design presents manufacturing challenges due to its numerous transistors and relies on innovative techniques for seamless operation. Hence, Nvidia’s manufacturing partner, Taiwan Semiconductor Manufacturing Co., will utilise its 4NP technique for production.
Blackwell will also enhance connectivity with other chips and improve AI data processing capabilities, marking the next step in Nvidia’s “superchip” lineup. Users will have the option to pair Blackwell with Nvidia’s Grace central processing unit and select from networking chips utilising proprietary InfiniBand or Ethernet protocols. Nvidia also plans to update its HGX server machines with the new chip.
Originating as a provider of graphics cards popular amongst gamers, Nvidia’s technology has been successful in various fields due to its ability to handle complex tasks in parallel. Blackwell aims to extend its reach beyond simple AI tasks, such as speech recognition, to more complex applications like generating three-dimensional videos.
Nvidia CEO Jensen Huang says robots are the future
Huang, while delivering his keynote, said that AI is the “driving force” as the economy goes through a change and the new Blackwell chips by Nvidia are the “engine to power this new industrial revolution.”
He then added that Nvidia is working with “the most dynamic companies in the world” and will recognise the potential of AI in every industry.
Huang also spilled the beans about Project Groot, which consists of a new computer based on Blackwell. He added that Project Groot will be available for humanoid robots manufacturers. It will enable robots to learn human-like movements and understand natural language by observing them. Huang mentioned that tackling these challenges was one of the most exciting tasks in AI at that time.
Huang ended the event on a creative note – asking two robots to join him on the stage. He revealed that these robots were trained using Nvidia’s simulation tools and said that in future, “everything that moves will be robotic.”
Source link
credite