True to form, NVIDIA CEO Jensen Huang delivered a stirring keynote address at GTC yesterday. Clad in his usual black leather jacket, he announced a million developments (or at least, it felt like it) in a 90-minute presentation. What he shared during that time is more than what many companies will announce over several years.
With little time to handle and still reeling from the content overload (six press releases and 27 blog posts), it is heartening that Huang took time to address the media in a question and answer session this morning. Here are seven highlights from the announcements:
- Accelerated computing. Sixty-five new and updated libraries with improved features and capabilities for data scientists, researchers, students and developers working on computing challenges. These include Morpheus cyber security library module and ReOpt AI software that optimises route planning, warehouse picking, fleet management and other logistics challenges to control rising delivery costs and improve efficiency.
- Triton AI inference platform. Triton is used by more than 25,000 customers worldwide, including Capital One, Microsoft, Samsung Medison, Siemens Energy, and Snap.
- Speech AI. NVIDIA Riva Custom Voice, a feature in NVIDIA Riva speech AI software, lets enterprises develop an expressive custom voice with Riva using a small amount of data in hours instead of weeks. According to Huang, this is an extremely important development as the way most people interact with computers is unstructured.
- Large language model (LLM). NVIDIA announced the NVIDIA NeMo Megatron framework for training language models with trillions of parameters and the Megatron 530B customisable LLM that can be trained for new domains and languages.
- Omniverse. The metaverse is probably the hottest topic today and NVIDIA Omniverse just got better with Omniverse Avatar for creating real-time conversational AI assistants and the Omniverse Replicator synthetic data generation engine that produces physically simulated synthetic data for training deep neural networks.
- Jetson AGX Orin Robotics Computer. It is claimed to be the world’s smallest, most powerful and energy-efficient AI supercomputer for robotics, autonomous machines, medical devices and other forms of embedded computing at the edge.
- Supercomputing. The NVIDIA Quantum-2 is the next generation of the InfiniBand networking platform that offers the performance, accessibility and security needed by cloud computing providers and supercomputing centres.
At the end of his keynote, Huang announced that NVIDIA will build Earth Two, a digital twin of the earth, to simulate and predict climate change.
“All the technologies we have invented up to this moment are needed to make Earth Two possible. I can’t imagine a greater and more important use,” he said.