NVIDIA
Latest in NVIDIA
At SIGGRAPH 2024, Mark Zuckerberg and Jensen Huang discussed AI assistants and open source, as Meta launched AI Studio for creating AI characters.
Speaking at SIGGRAPH 2024, NVIDIA CEO Jensen Huang said every job will have AI assistance and that accelerated computing saves energy.
NVIDIA has launched four NeMo Retriever NIM microservices that connect LLMs to business data for RAG, giving 30% fewer inaccurate answers.
Mistral AI and NVIDIA have released Mistral NeMo, a 12B multilingual model with a 128K context window, under the Apache 2.0 licence.
A study in Nature used GPU-powered supercomputer simulations of Ising spin glass to validate a possible route to commercial quantum annealing.
Mistral and NVIDIA released Mistral NeMo, a 12-billion-parameter model with a 128k context window under Apache 2.0, trained with quantisation awareness for FP8 inference.
Japan’s AIST is launching ABCI 3.0, an AI supercomputer using NVIDIA H200 GPUs, to boost national AI research and technological independence.
Toronto start-up Waabi is using generative AI and NVIDIA technology to develop self-driving trucks and plans fully driverless operations next year.
NVIDIA’s GauGAN, an early generative AI model, has evolved into the NVIDIA Canvas app, which turns rough sketches into realistic landscapes.
SyncTwin has built a digital twin app for factory planning using NVIDIA cuOpt, Omniverse and OpenUSD to optimise manufacturing processes and efficiency.
EvolutionaryScale, backed by NVIDIA and Amazon, has released ESM3, a generative AI model for protein design trained on about 2.8 billion protein sequences.
Taiwan’s leading medical centres and research institutes are using AI for uses such as colorectal cancer detection and autoimmune disease screening.
At COMPUTEX, NVIDIA set out its Blackwell architecture and partner systems, saying AI factories and data centres will drive the next industrial revolution.
Manufacturers can now build digital twins of factories to optimise layouts and processes without costly physical changes, NVIDIA’s Deepu Talla says.
Foxconn has built a digital twin of its new factory in Guadalajara, Mexico, so engineers can define processes and train robots in a virtual space.