News from the AI & ML world
DeeperML
Dean Takahashi@AI News | VentureBeat
//
Nvidia is significantly expanding its presence in the AI supercomputing and robotics landscape, announcing several key initiatives. The company is partnering with manufacturing leaders to establish AI supercomputer factories in the United States, a move aimed at solidifying its role in the AI infrastructure supply chain. This includes commissioning over a million square feet of manufacturing space in Arizona and Texas to build and test Nvidia Blackwell chips and AI supercomputers. Nvidia is also working with Amkor and SPIL for packaging and testing operations in Arizona.
Nvidia is making strides in long-context language models (LLMs) with the release of UltraLong-8B, a series of models designed to process extensive sequences of text, ranging from 1 million to 4 million tokens. This breakthrough addresses the challenge of limited context windows in LLMs, enabling the models to handle longer documents and videos without sacrificing performance. Researchers from NVIDIA and UIUC have developed a technique to extend context window limits in LLMs, successfully increasing Llama 3.1’s context window from 128k to 4M tokens.
Further demonstrating its commitment to AI and robotics, Nvidia is preparing to launch Jetson Thor in the first half of 2025, a compact computing platform specifically designed for humanoid robots. This move positions Nvidia as a comprehensive technology provider for the "physical AI" domain, offering both hardware and software solutions for robot training, simulation, deployment, and optimization. The launch of Jetson Thor underscores Nvidia's strategy to capture a significant portion of the rapidly growing robotics market, projected to reach between $190 billion and $910 billion by 2040.
ImgSrc: venturebeat.com
References :
Nvidia is making strides in long-context language models (LLMs) with the release of UltraLong-8B, a series of models designed to process extensive sequences of text, ranging from 1 million to 4 million tokens. This breakthrough addresses the challenge of limited context windows in LLMs, enabling the models to handle longer documents and videos without sacrificing performance. Researchers from NVIDIA and UIUC have developed a technique to extend context window limits in LLMs, successfully increasing Llama 3.1’s context window from 128k to 4M tokens.
Further demonstrating its commitment to AI and robotics, Nvidia is preparing to launch Jetson Thor in the first half of 2025, a compact computing platform specifically designed for humanoid robots. This move positions Nvidia as a comprehensive technology provider for the "physical AI" domain, offering both hardware and software solutions for robot training, simulation, deployment, and optimization. The launch of Jetson Thor underscores Nvidia's strategy to capture a significant portion of the rapidly growing robotics market, projected to reach between $190 billion and $910 billion by 2040.
ImgSrc: venturebeat.com
Share:
References :
- AI News | VentureBeat: Nvidia pledges to build its own factories in the U.S. for the first time to make AI supercomputers
- www.marktechpost.com: NVIDIA AI Releases UltraLong-8B: A Series of Ultra-Long Context Language Models Designed to Process Extensive Sequences of Text (up to 1M, 2M, and 4M tokens)
- analyticsindiamag.com: Thanks to NVIDIA, Llama 3.1’s Context Window Went Up From 128k to 4M
- www.cnbc.com: Nvidia to mass produce AI supercomputers in Texas as part of $500 billion U.S. push
- analyticsindiamag.com: NVIDIA to Manufacture First American-Made AI Supercomputers
- the-decoder.com: Nvidia shifts AI production to US amid changing trade landscape
Classification:
- HashTags: #NVIDIA #AISupercomputers #LongContextLLMs
- Company: NVIDIA
- Target: AI Researchers, Enterprises
- Product: Jetson Thor, UltraLong-8B
- Feature: AI Supercomputing
- Type: AI
- Severity: Informative