AI/ML news, top picks, and generated innovation digests.
116 articles tagged with this keyword, sorted by most recent first.
Anthropic’s Claude models in Microsoft Foundry — hosted on Microsoft Azure and running on NVIDIA GB300 Blackwell Ultra GPUs — are now generally available, giving Azure-native enterprises a powerful new way to build autonomous and domain-specific AI agents. As agentic AI continues to drive enterprise innovation and becomes more autonomous, organizations need access to computing […]
Ornith-1.0: Self-Scaffolding LLMs for Agentic Coding This is an interesting new open weights (MIT licensed) model, the first model release from DeepReinforce. [...] with variants including 9B Dense, 31B Dense, 35B MoE, and 397B MoE. Built on top of pretrained Gemma 4 and Qwen 3.5, it achieves state-of-the-art performance among open-source models of comparable size on coding benchmarks. As far as I can tell the licenses of those underlying models is compatible with being used in this way - Gemma 4 is Apache 2.0 licensed (and not bound by the janky additional Gemma Terms of Use that afflicted the previous Gemma models) and Qwen 3.5 is Apache 2.0 licensed as well. I've been running the model using LM Studio and the ornith-1.0-35b-Q4_K_M.gguf (20GB) GGUF, hooked up to Pi . Initial impressions are very good - it seems to be able to run the agent harness over many tool calls in a proficient way. Here's a terminal session where I asked it to "find the code that decodes the actor cookie" and then "find the code that opens the insert dialog when thebutton is clicked" against a Datasette checkout, which it handled with ease. I also had it draw this pelican , which came out at 103 tokens/second: It's a little bit mangled but the pelican is clearly a pelican. I couldn't find much information about DeepReinforce themselves. The earliest paper I could find from the was CUDA-L1: Improving CUDA Optimization via Contrastive Reinforcement Learning from June 2025. Tags: ai , generative-ai , lo…
Showcasing the importance of open source innovation in American AI, Palantir’s new intelligent engine — introduced today — uses NVIDIA Nemotron open models to serve the needs of U.S. government agencies. Open source software has long been a pillar of U.S. technology leadership. In 1969, DARPA connected four university computers — from UCLA, Stanford, UCSB […]
I like the direction of this article, though I think it could go deeper into real-world applications. The practical side is where most users struggle..
Flexion Robotics, a startup founded by ex-Nvidia engineers, has a clever way of training robots to do useful work.
Adtech’s agentic moment arrived at Cannes Lions 2026. Major players, including Nvidia, Yahoo, Palantir, Pinterest, and Fox, among others, unveiled foundational infrastructure that allows AI agents to connect with each other to perform marketing tasks previously done by humans.
Nvidia's AI chip sales in China stall, as local chipmakers like Huawei take the lead
Tech and AI stocks now make up as much as 12% of most balanced superannuation funds, experts say Follow our Australia news live blog for latest updates Get our breaking news email , free app or daily news podcast Artificial intelligence and technology stocks have become a driving force on Wall Street and, unbeknownst to most Australians, a growing part of their retirement savings . The so-called “magnificent seven” – chip maker Nvidia, Google owner Alphabet, Apple, Microsoft, Amazon, Facebook owner Meta and Tesla – are, for better or worse, increasingly part of the portfolios offered by superannuation funds. Continue reading...
Eager to find more public AI-related companies that may do as well as Nvidia, Wall Street investors think they've found a winner with Micron.
The discussion about nvidia’s global context vit achieves sota performance on cv tasks without expensive computation raises some really valid points. This perspective is refreshing. gemini music
The NVIDIA Nemotron Model Reasoning Challenge on Kaggle on Kaggle brought together 5,000+ participants across 4,000+ teams to explore how builders can improve reasoning accuracy using open models, shared benchmarks, and reproducible workflows. Join NVIDIA Kaggle Grandmasters and challenge winners for a live discussion on the techniques that moved the leaderboard, from verified reasoning traces and token-aware prompts to solver-driven data pipelines, targeted fine-tuning, and better validation. We’ll also highlight community discoveries from notebooks and discussion threads that helped teams debug, iterate, and improve. What you'll learn: How verified reasoning traces can improve training signal How to design prompts and traces around token budget How solvers and tools can create better reasoning data How to compare techniques across task types, not just aggregate scores What open models like Nemotron make possible for community experimentation Experimenting with Nemotron reasoning models or working on your own benchmarks? Bring your questions live — and we will answer them in real time.
In this tutorial, we work with NVIDIA's Open-SWE-Traces dataset to study agentic software-engineering trajectories for fine-tuning. We stream the data directly from Hugging Face, so we can process it efficiently in Google Colab without downloading everything locally. We normalize multi-turn agent conversations, parse final code patches, and build an analysis DataFrame covering trajectory length, tool usage, patch size, language distribution, and resolution outcomes. We then curate a supervised fine-tuning subset using success labels, token limits, language filters, and patch availability. The post Building Supervised Fine-Tuning Data from NVIDIA Open-SWE-Traces: Trajectory Parsing, Patch Analysis, Token Budgets, and Tool-Use Metrics appeared first on MarkTechPost .
Nvidia has dominated the AI chip market for years, but the era of total dependence might be ending. OpenAI just shared its plans to spice things up with Jalapeño, its custom inference chip built with Broadcom, joining Google, Apple, and SpaceX in a growing list of companies building their way out of single-supplier risk. The goal is less of a […]
This post shows you how to configure training jobs on Amazon SageMaker AI to get the most out of Blackwell’s architecture on AWS. You learn how to select batch sizes and sequence lengths that take advantage of Blackwell’s expanded memory, choose the right precision format for your model size (1B to 64B parameters), and apply activation checkpointing strategically. By the end, you have a practical framework for tuning your training configuration and launching distributed training jobs on P6-B200 instances.
Let’s walk through the NVIDIA cuFOLIO Developer Example. This open source, customizable notebook enables GPU accelerated portfolio optimization by constructing an optimal portfolio from the S&P 500 universe and then backtesting against customizable parameters and portfolios. ➡️ Start now: https://build.nvidia.com/nvidia/quantitative-portfolio-optimization 📥 Code: https://github.com/NVIDIA-AI-Blueprints/quantitative-portfolio-optimization/ 📝 Tech blog: https://developer.nvidia.com/blog/accelerate-large-linear-programming-problems-with-nvidia-cuopt 00:00 Interactive Backtesting Intro 00:11 Quantitative Portfolio Optimization 00:26 Deploy on Cloud (Brev) 00:57 Launchable Setup 01:50 Github 01:57 Run Notebook 02:42 2. CVaR Formulation 03:00 3. Data and Model Setup 04:26 4. Solve CVaR Optimization 07:15 5. Backtest Portfolio 08:09 6. GPU v CPU 09:40 7. Appendix 10:05 Outro #quantfinance #portfoliooptimization #algorithmictrading
NVIDIA Spark Hack Toronto brought developers together for a weekend challenge: build an agentic application that runs locally on DGX Spark using open models and Toronto Open Data. Teams tackled everything from small business forecasting and dementia care to city-scale traffic simulation — all on an ASUS Ascent GX10 powered by the NVIDIA GB10 Grace Blackwell Superchip. Cracked City took home Best Use of Nemotron for turning road and sidewalk damage reporting into a single step. Upload a photo, and their system analyzes the damage, estimates severity, and auto-generates a Toronto 311 report using image analysis, speech recognition, and NVIDIA Nemotron — all running locally on DGX Spark. Join us live to see a demo and chat with the winners. Bring your questions about building with Nemotron and local AI compute.
OpenAI’s financial trajectory hinges heavily on infrastructure costs, a reality that drove the development of the new custom OpenAI Jalapeño chip. Developed in collaboration with Broadcom, the application-specific integrated circuit (ASIC) represents a direct attempt to mitigate the heavy capital expenditure associated with third-party hardware. While Nvidia currently commands an estimated 75% profit margin on […] The post The math behind the OpenAI Jalapeño chip appeared first on AI News .
The rising popularity of AI, Edge, and Telecommunications workloads on Kubernetes has led to new requirements for hardware management. We now need hardware specification beyond CPU time and memory allocations. This includes allocating GPUs, TPUs, network interfaces, and other hardware, sometimes after pod start and occasionally through time-sharing. Efficiently managing this specialized hardware is the mission of the Device Management Working Group . Their cornerstone project, Dynamic Resource Allocation (DRA) , recently graduated to GA, marking a fundamental shift in how the project handles hardware-intensive workloads at scale. In this spotlight, we sit down with working group chairs Kevin Klues , Patrick Ohly , and John Belamaric to discuss the limitations of the legacy device model, the NP-hard challenges of scheduling, and how they’re building a more programmable, hardware-aware future for Kubernetes. Introducing Device Management Natalie Fisher: Can you introduce yourself, your role, and how you got involved in the Device Management Working Group? Kevin Klues: My name is Kevin Klues. I am a Distinguished Engineer at NVIDIA. I have been a co-chair of the device management working group since its inception at Kubecon EU 2024. I have also been involved with DRA (the working group's primary deliverable) since its inception in 2019 / 2020. I have also been a kubelet maintainer since 2019, with a focus on its device manager, CPU manager, and topology manager subcomponents. T…
PLUS: Trump's quantum deadline, SpaceX's $6.3B deal, NVIDIA robots
NVIDIA has released the full Nemotron 3 open model family — Ultra, Super, Nano, and Nano Omni. This office hours session covers each model in the series, and any questions you have about Nemotron 3 in general — what it's built for, when to use it, and what's available in open weights, training datasets, and fine-tuning recipes. What we'll cover: - Nemotron 3 Ultra — 550B MoE frontier reasoning model for long-running autonomous agents: 5x faster inference, up to 30% lower cost, hybrid Mamba-Transformer architecture, and MOPD training for consistent performance across agent harnesses - Nemotron 3 Super — mid-range 120B model targeting enterprise applications that need strong reasoning for multi-agent applications - Nemotron 3 Nano — 30B MoE with 3B active parameters, built for high-volume execution, highly accurate sub-agent accomplishing targeted tasks - Nemotron 3 Nano Omni — multimodal (text, image, audio, video) model purpose-built for targeted specialized agentic tasks - Open weights, training datasets, and fine-tuning recipes — what's available across the family and how to customize for your domain Building with or evaluating the Nemotron 3 family? Bring your questions — whether you're choosing between models, fine-tuning for your domain, or deploying at scale, the team will answer them live.
The post This Bengaluru startup isn’t building faster chips than Nvidia. It’s building cheaper AI appeared first on The Ken .
AI agents place new demands on inference infrastructure. Unlike a single chatbot response, an agentic workflow can involve many LLM calls, tool calls, long context windows, and repeated cache reuse across a task. NVIDIA Blackwell is designed to handle these production-scale agent workloads with high throughput, low latency, and improved energy efficiency. This livestream explains how NVIDIA Blackwell helps developers scale AI agents in production, using AgentPerf results as one example of its performance on real-world coding-agent workloads. We’ll also cover how NVIDIA Dynamo adds software-level optimizations for routing, scheduling, and KV cache management. What you’ll learn: Why AI agents require different infrastructure than standard chat applications. How NVIDIA Blackwell improves throughput and efficiency for concurrent agent workloads. What AgentPerf results show about Blackwell performance on realistic agentic coding tasks. How Dynamo optimizes inference with agent-aware routing, scheduling, and KV cache reuse. What developers should consider when deploying AI agents at production scale.
Building AI systems at scale is demanding, requiring low-latency inference, fast vector search, strong GPU price-performance and infrastructure that can grow without multiplying operational complexity. NVIDIA’s latest work with Amazon Web Services (AWS) addresses each of those constraints. Across Amazon OpenSearch and Amazon EC2, NVIDIA AI infrastructure is giving enterprises more practical paths to deploy […]
Artificial intelligence played a prominent role at this week’s Bio International Convention in San Diego, the largest biotech event with vendors spanning the full ecosystem of companies in this industry. Today in a special address, Kimberly Powell (pictured), vice president and general manager of healthcare and life sciences at Nvidia Corp., made the case that agentic AI […] The post Nvidia bets on agentic AI to turbocharge biotech discovery appeared first on SiliconANGLE .
Editor’s note: This post is part of the Nemotron Labs blog series, which explores how the latest open models, datasets and training techniques help businesses build specialized AI systems and applications on NVIDIA platforms. Each post highlights practical ways to use an open stack to deliver real value in production — from transparent research copilots […]
NVIDIA technologies power more than 400 of the world’s 500 fastest supercomputers — 81% of the TOP500 — according to the latest rankings released this week at the ISC High Performance conference in Hamburg, Germany.
Telecom operators have seen remarkable returns from using generative AI to automate network management, customer care and back-office operations. Most of that impact has been task‑based: automation that speeds up predetermined steps while people manually correlate insights and direct next steps. Automation is no longer the finish line — it’s the launchpad to autonomy. The […]
This morning on Hacker News I saw Moebius: 0.2B Lightweight Image Inpainting Framework with 10B-Level Performance , describing a small but effective inpainting model - a model where you can mark regions of an image to remove and the model imagines what should fill the space. The released model required PyTorch and NVIDIA CUDA , but since it described itself as 0.2B I decided to try and get it running using WebGPU in a browser. TL;DR: I got it working, and you can try the demo at simonw.github.io/moebius-web/ . Read on for the details. The finished tool Here's a video demo of the finished tool: You can open any image in it (non-square images get letterboxed), highlight areas to remove, click the "Run inpaint" button and wait for the model to do its magic. A parallel agent side-project My main project for today was landing a major feature in Datasette: a UI for creating and altering tables, as a follow-up to the insert and edit rows feature I released last week. I was working on that in Codex Desktop (here's the PR ) and often found myself spending 5-10 minutes spinning my fingers waiting for it to complete a mid-sized refactor or add the finishing touches to a change to the UI. (An amusing thing about coding agents is that the harder a problem is the more time you have to get distracted while you wait for them to finish crunching!) So I decided to spin up Claude Code in a terminal window and see how far I could get at porting Moebius to the web. Some agentic research to kick…
JUPITER, Europe’s first exascale supercomputer at Germany’s Forschungszentrum Jülich, runs on NVIDIA Grace Hopper Superchips and NVIDIA Quantum-X800 InfiniBand networking — and it’s had a busy year. As the international supercomputing community gathers at ISC in Hamburg this week, four projects running on JUPITER point to what exascale computing can actually do: map the human […]
For the past two years, the U.S. National Science Foundation’s National Artificial Intelligence Research Resource (NAIRR) pilot program has driven innovative research across the U.S. for over 700 projects — spanning protein prediction and infectious disease outbreak management. NVIDIA contributed to the NAIRR pilot through a cloud-based resource that gives researchers dedicated access to a […]
Mission, Vision and Veritas — new Los Alamos National Laboratory (LANL) supercomputers to be built with HPE and NVIDIA — are tapping NVIDIA Vera CPUs to accelerate scientific discovery, unlocking agentic AI for science. The supercomputers will use the HPE Cray Supercomputing GX5000 architecture with the NVIDIA Vera Rubin platform, combining NVIDIA Vera CPUs, NVIDIA […]
At the ISC conference running in Hamburg this week, NVIDIA is introducing new software that speeds AI for science, from chemistry and materials discovery to the search for dark matter. The NVIDIA DAQIRI library and new NVIDIA ALCHEMI NIM microservices — as well as the NVIDIA cuPhoton reference code, coming soon — turn work that […]
The next era of AI will not be defined by compute alone. Its growth will be determined by energy. As accelerated computing scales across AI factories, agentic AI, industrial AI, edge computing and physical AI — including robotics and autonomous systems — global electricity demand is rising at unprecedented speed. In many regions, expanding grid […]
Hot tubs sit at about 38 to 40 degrees Celsius, warm enough that most people can only soak for about 15 minutes. NVIDIA’s newest AI servers can run their cooling liquid even hotter — up to 45 degrees Celsius, or 113 degrees Fahrenheit. That higher temperature limit is precisely what makes them more energy efficient. […]
In a consequential grid infrastructure decision, the Federal Energy Regulatory Commission (FERC) today issued a major milestone on large-load interconnection impacting how those building AI factories, semiconductor fabrication support systems and advanced manufacturing facilities can connect to the grid. In the era of AI, which NVIDIA founder and CEO Jensen Huang has described as a […]
The digital era gave the advertising and marketing industry speed; the AI era is giving it autonomous operations. For companies building next-generation technologies for advertising and marketing, the question is no longer whether to adopt AI but whether their infrastructure can support it at the speed and scale the industry demands. At Cannes Lions, running […]
A year ago at NVIDIA GTC Paris at VivaTech, France laid out plans to advance local AI — from new AI factories and national compute capacity to open frontier models and industrial platforms. Now, that AI infrastructure is coming online. AI agents are running in production, startups are deploying applications and the French AI ecosystem […]
Nvidia's self-improvement program for robots enlists teams of AI coding agents.
Nemotron 3 Ultra is NVIDIA's latest frontier-intelligence open model — 5x faster inference, up to 30% lower cost, and fully open: weights, training datasets, and fine-tuning recipes included. In this livestream, we're joined by Nathan Lambert, ML researcher and open model advocate, to dig into what Ultra means for developers building on open models today. We'll cover what sets Ultra apart technically — the hybrid Mamba-Transformer backbone, Multi-Teacher On-Policy Distillation (MOPD), and how it fits into a system-of-models pattern. Nathan brings a researcher's perspective on post-training for agentic systems, and we'll get into where the open frontier model landscape is heading and what it takes to build models worth building on. What you'll learn: - How Ultra's post-training approach compares to what the open model ecosystem has seen at scale - What the hybrid Mamba-Transformer architecture means for long-context, multi-turn agent workflows - How open weights, datasets, and recipes enable domain-specific fine-tuning from day one - Where open frontier models are heading for agentic applications — and what tradeoffs matter most Have questions about Ultra, post-training, or the open model landscape? Drop them live — Nathan and the team will answer them in real time.
NVIDIA XR AI is now available in public beta, giving developers a framework for building multimodal AI agents for AR glasses and XR devices.
Enterprises are moving agentic AI from proof of concept to production — and the next generation of AI factories are built for the era of agents. At HPE Discover Las Vegas, running through Thursday, June 18, NVIDIA and HPE are expanding the HPE AI Factory with NVIDIA, including NVIDIA Vera CPU and NVIDIA Agent Toolkit […]
In this session we will focus on how to bring VLM/VLA models to power real-world physical AI applications. We will focus on how to utilize SOTA of VLM (gemma 4) and or GR00T model for performing different pick and place tasks and orchestrate the outputs to control the robots using ROS 2 framework. You will learn how to bring vision-language models into real-world physical AI applications — from model selection to robot control. We'll cover: Choosing the right model for robotics — learn when to use a state-of-the-art VLM like Gemma 4 versus a specialized model like NVIDIA GR00T, and how runtime, throughput, and task requirements shape that decision. VLMs and VLAs in action — see how vision-language and vision-language-action models are applied to real manipulation tasks like pick and place, and what makes them viable for physical AI. Connecting model outputs to robot control — understand how to orchestrate model outputs through the ROS 2 framework to drive real robot behavior. Hands-on hardware demo — walk through a live example using the SO-101 or reBot Arm, putting everything together from model inference to physical actuation.
This session moves from running a local model to running a local autonomous agent. OpenClaw is a fully local AI assistant that runs on Jetson and connects to chat workflows, browser-based tools, and multi-step tasks. NemoClaw extends this with sandboxing, onboarding, inference routing, and policy controls for safer and more structured agent deployments. We'll show what changes when an AI system can take actions, use tools, and run privately on your own hardware — 24/7, at home, on the edge. Use cases include building dynamic browser-based games, prototyping smart computer vision apps, and running long research tasks without a cloud dependency. You will learn how to move from running a local model to running a fully local autonomous agent on NVIDIA Jetson. We'll cover: Building a local assistant with OpenClaw — extend the Episode 1 baseline into a full local assistant architecture that connects to chat workflows, browser-based tools, and multi-step tasks — running privately on your own hardware, 24/7. NVIDIA Orin Nano vs. AGX Orin vs. Thor — compare hardware paths side by side so you can make the right choice for your deployment constraints and performance needs. Why tool-calling models matter — see what changes when an AI system can take actions, use tools, and run autonomously, and what breaks when your model can't do it reliably. Safer local agents with NemoClaw — go further with sandboxing, onboarding, inference routing, and policy controls that make local agent deploymen…
This opening session builds the foundation for running popular OSS models such as Gemma, Qwen directly on Jetson — no cloud required. We cover when to use Ollama for rapid local prototyping versus vLLM for higher-throughput serving, show how the same workflow applies to both power different OSS models, and walk through the real decisions behind model choice, containers, quantization, and performance tuning on edge hardware. We close with a teaser of OpenClaw and a bonus take-home challenge to kick off community building. You will learn how to deploy open-source AI models on NVIDIA Jetson — no cloud required, from first launch to production-ready serving. We'll cover: Getting models running on NVIDIA Jetson — spin up popular OSS models (open-source large language models (LLMs) like Gemma and Qwen (LLMs and VLMs) using Ollama or vLLM on Jetson hardware and verify they're working end-to-end. Choosing the right inference engine — understand the practical tradeoffs between Ollama for rapid local prototyping, vLLM for higher-throughput serving, and llama.cpp, so you can pick the right tool for your use case. NVIDIA Jetson-specific serving strategies — walk through the real decisions behind model choice, containers, and performance tuning tailored for Orin and Thor, including what works, what doesn't, and why. Performance fundamentals — get introduced to quantization and speculative decoding: what they are, how they work, and when to reach for them on edge hardware. Real-world appl…
Starting with JetPack 7.2, NVIDIA officially supports the Yocto Project on Jetson. But the story began years earlier with meta-tegra, the community project created and maintained by Matt Madison that brought Yocto Project to Jetson and became a trusted foundation for embedded Linux developers. Join Matt Madison and NVIDIA experts as they discuss the origins of meta-tegra, the journey from community-driven project to official NVIDIA support, and what this milestone means for developers building products with Jetson. What you'll learn: - The origin of meta-tegra and the problem it was built to solve - What NVIDIA's commitment to OE4T means in practice for the codebase and community - Whether the Yocto Project is right for your Jetson product—and how to get started Join us live, bring your questions, and hear the story directly from the people who built it.
❤️ Check out Lambda here and sign up for their GPU Cloud: https://lambda.ai/papers 📝 The Nemotron 3 Ultra paper is available here: https://research.nvidia.com/labs/nemotron/Nemotron-3-Ultra/ Free Rendering course and source code: https://users.cg.tuwien.ac.at/zsolnai/gfx/rendering-course/ 🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible: Adam Bridges, Benji Rabhan, B Shang, Cameron Navor, Charles Ian Norman Venn, Christian Ahlin, Eric T, Fred R, Gordon Child, Juan Benet, Michael Tedder, Owen Skarpness, Richard Sundvall, Ryan Stankye, Shawn Becker, Steef, Taras Bobrovytsky, Tazaur Sagenclaw, Tybie Fitzhugh, Ueli Gallizzi Thumbnail design: https://felicia.hu #nvidia
NVIDIA Spark Hack Toronto brought developers together for a weekend challenge: build an agentic application that runs locally on DGX Spark using open models and Toronto Open Data. Teams tackled everything from small business forecasting and dementia care to city-scale traffic simulation — all on an ASUS Ascent GX10 powered by the NVIDIA GB10 Grace Blackwell Superchip. Belong won the Public Services track with an AI companion designed for people living with dementia and their caregivers. Running entirely on DGX Spark, it helps users recognize family members, remember appointments, and find local services — keeping all conversations and memory data on-device using Nemotron, speech, and retrieval systems. CityFlow won the Economic Systems track with a real-time intelligence platform for small businesses. By combining transit disruptions, road closures, weather, events, and other Toronto data sources, it helps business owners answer practical questions like "Am I properly staffed for Friday night?" — using Nemotron to generate recommendations and NVIDIA cuOpt to turn forecasts into actual staffing plans. Join us live to see demos and chat with both teams. Bring your questions about building with local AI compute and Toronto Open Data.
Join NVIDIA for a live demonstration of how developers can generate synthetic data for physical AI using NVIDIA Brev Launchables and agent skills. Building synthetic data pipelines for robotics, digital twins, and autonomous systems often requires configuring GPU infrastructure, simulation environments, notebooks, and orchestration tools before meaningful work can begin. In this livestream, we'll show how NVIDIA Brev Launchables and agent skills simplify that process by packaging these components into ready-to-run workflows that help developers move from setup to data generation faster. In this livestream, you'll learn how to: - Launch preconfigured Physical AI development environments - Generate synthetic data using AI-powered workflows - Accelerate robotics, simulation, and digital twin development - Scale from individual tasks to larger synthetic data pipelines - Integrate data generation workflows into broader Physical AI ecosystems Through live, hands-on demonstrations, we'll show how developers can streamline synthetic data creation and reduce the complexity of building Physical AI workflows. Whether you're building robots, training computer vision models, creating digital twins, developing autonomous systems, or exploring Physical AI applications, this session provides a practical introduction to synthetic data generation with NVIDIA Brev Launchables and agent skills. -------------------------------- 📓 Resources Launchable: - Nurec: https://brev.nvidia.com/launchable/…
GPUs aren’t just for LLMs; they are accelerating life saving discoveries in tabular data science. On the next Google Cloud Live livestream, join experts from Google Cloud and NVIDIA for a live, end-to-end breakdown of GPU-accelerated virtual drug screening. Hosted by Tilde, alongside Jeff Nelson, William Hill, and Dr. Saee Paliwal, discover how to take molecular predictions from pipeline to production. Watch along and learn about: Interactive live demo: Drop everyday compounds in the chat and watch our web app predict lung cancer (EGFR) binding likelihood in seconds. GPU-accelerated pipelines: Learn how to get 20x-45x training speedups using cuDF and cuML without rewriting your pandas or scikit-learn code. Stop waiting on CPU bottlenecks and learn how to virtualize screening at the trillion molecule scales. Speakers: Tilde Thurium, Jeff Nelson, William Hill, Saee Paliwal Products Mentioned: GPU, NVIDIA, Google Cloud
I'm trying to train a GAN model, but its results are very bad. The Generator doesn't seem to work. Can someone suggest how this can be improved? What is the best way to remember the code? Can this also be done using a builtin library? import os import torch import torchvision import torch.nn as nn import torch.optim as optim import torch.nn.functional as F import torchvision.datasets as datasets import torchvision.transforms as transforms from torch.utils.data import DataLoader import matplotlib.pyplot as plt import numpy as np random_seed = 42 torch.manual_seed(random_seed) BATCH_SIZE = 128 AVAIL_GPUS = min(1, torch.cuda.device_count()) DEVICE = torch.device("cuda" if AVAIL_GPUS else "cpu") LATENT_DIM = 100 transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize((0.5,), (0.5,)) # scale to [-1, 1] for tanh output ]) dataset = datasets.MNIST(root="./data", train=True, download=True, transform=transform) dataloader = DataLoader(dataset, batch_size=BATCH_SIZE, shuffle=True, num_workers=2, drop_last=True) class Generator(nn.Module): def __init__(self, latent_dim=100, img_channels=1, feature_maps=64): super().__init__() self.net = nn.Sequential( nn.ConvTranspose2d(latent_dim, feature_maps * 4, kernel_size=7, stride=1, padding=0, bias=False), nn.BatchNorm2d(feature_maps * 4), nn.ReLU(True), nn.ConvTranspose2d(feature_maps * 4, feature_maps * 2, kernel_size=4, stride=2, padding=1, bias=False), nn.BatchNorm2d(feature_maps * 2), nn.ReLU(True), nn.ConvTranspose2d(…
TL;DR Helion kernels were integrated into vLLM for FP8 inference using Qwen3 models and evaluated across NVIDIA H100 and B200 GPUs. The experiments show that Helion provides a productive PyTorch-native...
At Computex 2026, an annual computer trade show held in Taipei, Taiwan, Nvidia made a long anticipated announcement—a version of the company’s Blackwell GB10 superchip for Windows PCs, called RTX Spark. Originally rumored to launch in 2025 , it was finally introduced at this year’s show. It came with full support from Microsoft, which announced two new devices powered by RTX Spark: the Surface Laptop Ultra and the Surface RTX Spark Dev Box . Asus, Dell, Lenovo, HP, and MSI also announced Windows PCs with RTX Spark. If this is triggering déjà vu, that’s for good reason. In June 2024, Qualcomm and Microsoft partnered to launch AI-focused Copilot+ PCs. Qualcomm’s Arm-based chips provided an alternative to x86-based chips from AMD and Intel used across dozens of budget and mid-range Windows laptops. It was met with mixed commercial success, however, and Intel remains the dominant supplier of chips for Windows laptops. But that doesn’t mean RTX Spark will follow the same path, as Nvidia’s involvement is an important part of the equation. “Nvidia just has more clout and more industry weight to push and make things happen that Qualcomm couldn’t do early on, and that even Microsoft struggled with,” says Ryan Shrout , president at Signal65 , a third-party testing firm. “They can get game developers on board and get software developers in the emerging AI space to pay attention.” What is RTX Spark? At its core, RTX Spark is an iteration of the hardware found in the DGX Spark mini-works…
Short note on Nemotron 3 Ultra, NVIDIA's 550B total and 55B active hybrid Mamba-Transformer Latent MoE model.
Quantum computers promise to one day solve problems beyond the most powerful supercomputers imaginable. But it’s often underappreciated how much classical computing it takes just to operate these machines. As qubit counts rise, innovations in this supporting infrastructure will be essential if they’re to live up to their promise. To prepare for the scale of quantum computers the industry is working toward, many companies are also gearing up the classical hardware, and software, required to support them. In April, Nvidia announced new AI-based software to accelerate the classical tasks that enable quantum computers. Sydney-based quantum software company Q-CTRL has developed an automatic calibration algorithm for quantum computers, and is now leveraging Nvidia’s agent-based system. Other companies, including IBM Quantum , Cambridge, England–based Riverlane , which develops quantum-error correction, and Google Quantum AI , are developing similar tools. The Role of Classical in Quantum Digital computer chips are marvels of engineering, operating flawlessly out of the box and capable of trillions of operations without error. The quantum bits, or qubits, at the heart of a quantum computer, by contrast, are temperamental and unreliable, requiring regular calibration and complex error-correcting schemes to keep them on track. Calibration and error-correction are fundamentally classical, not quantum, problems, and they require dedicated classical hardware to solve. As quantum compute…
From a working group proposal in 2023 to a 40+ member industry effort with native support in PyTorch, NVIDIA NeMo, ASTRA-sim, and commercial simulation and emulation tools — MLCommons Chakra is reshaping how the AI systems community studies, reproduces, and co-designs training and inference platforms. The post Chakra Comes of Age: A Standardized Trace Ecosystem for AI Systems Benchmarking and Co-design appeared first on MLCommons .
NVIDIA and Microsoft birthed a new computer
Memory is arguably the most serious constraint on modern AI large language models (LLMs). According to one influential paper , LLM token generation is an inherently memory-bound task, meaning the rate at which models output text is limited by how quickly data can be read in from memory. The severity of this bottleneck grows with model size. This creates a “memory wall” that holds back LLM inference performance. AI hardware startup Majestic Labs is taking a direct—and comprehensive—approach to solving this problem. It’s developing a new AI server, Prometheus, with up to 128 terabytes of memory. That’s over 60 times more than Nvidia’s DGX B300 server , a cutting-edge AI processing rack. Sha Rabii , co-founder and president of Majestic Labs, believes that this drastic increase in memory will provide his company an edge. While he acknowledges that “Nvidia’s done a phenomenal job creating a system that can scale out,” he argues that it becomes less economical as models grow and “ends up greatly over-provisioning on compute and starving on memory.” DRAM-Centric Architecture for LLM Memory Majestic Labs plans to surmount the “memory wall” with an architecture that fundamentally differs from competitors’. Nvidia’s current servers have fast high-bandwidth memory (HBM), which is typically used to read in an LLM’s model weights. In addition, there’s an often larger but slower pool of dynamic random access memory (DRAM), which handles LLM and server overhead. Majestic instead goes all i…
Anthropic confidentially filed a draft S-1 with the SEC today for a proposed public offering. The company also shipped Claude Opus 4.8 last week with a 4x code-reliability gain. NVIDIA used GTC Taipei to open Cosmos 3, ramp Vera Rubin into production, and put a 1-petaflop AI box on developer laptops. Google retires Gemini 2.0 Flash today. California's SB 867 — banning AI companion chatbots in children's toys — cleared the Senate; Illinois's data-center regulation stalled in committee. The labs sprint. The states crawl.
The healthcare industry is ground zero for AI companies and the rollout of their products: Microsoft tells us that AI is better than doctors at diagnosing complex medical conditions. Nvidia claims that its chatbot, a partnership with the startup Hippocratic AI, can outperform nurses on detecting over the counter drug toxicities. AI firms suggest that […] The post Expanding our AI and Healthcare Portfolio appeared first on AI Now Institute .
❤️ Check out Lambda here and sign up for their GPU Cloud: https://lambda.ai/papers 📝 The paper is available here: https://arxiv.org/abs/2604.24954 https://developer.nvidia.com/blog/nvidia-nemotron-3-nano-omni-powers-multimodal-agent-reasoning-in-a-single-efficient-open-model/ https://huggingface.co/blog/nvidia/nemotron-3-nano-omni-multimodal-intelligence Our Patreon if you wish to support us: https://www.patreon.com/TwoMinutePapers 🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible: Adam Bridges, Benji Rabhan, B Shang, Cameron Navor, Charles Ian Norman Venn, Christian Ahlin, Eric T, Fred R, Gordon Child, Juan Benet, Michael Tedder, Owen Skarpness, Richard Sundvall, Ryan Stankye, Shawn Becker, Steef, Taras Bobrovytsky, Tazaur Sagenclaw, Tybie Fitzhugh, Ueli Gallizzi My research: https://cg.tuwien.ac.at/~zsolnai/ Thumbnail design: https://felicia.hu #nvidia
Jensen Huang is the co-founder and CEO of NVIDIA, the world’s most valuable company and the engine powering the AI computing revolution. Thank you for listening ❤ Check out our sponsors: https://lexfridman.com/sponsors/ep494-sc See below for timestamps, transcript, and to give feedback, submit questions, contact Lex, etc. Transcript: https://lexfridman.com/jensen-huang-transcript CONTACT LEX: Feedback – give feedback to Lex: https://lexfridman.com/survey AMA – submit questions, videos or call-in: https://lexfridman.com/ama Hiring – join our team: https://lexfridman.com/hiring Other – other ways to get in touch: https://lexfridman.com/contact EPISODE LINKS: NVIDIA: https://nvidia.com NVIDIA on X: https://x.com/nvidia NVIDIA AI on X: https://x.com/NVIDIAAI NVIDIA on YouTube: https://youtube.com/@nvidia NVIDIA on Instagram:
This is a transcript of Lex Fridman Podcast #494 with Jensen Huang. The timestamps in the transcript are clickable links that take you directly to that point in the main video. Please note that the transcript is human generated, and may have errors. Here are some useful links: Go back to this episode’s main page Watch the full YouTube version of the podcast Table of Contents Here are the loose “chapters” in the conversation. Click link to jump approximately to that part in the transcript: 0:00 – Introduction 0:33 – Extreme co-design and rack-scale engineering 3:18 – How Jensen runs
Robert Lange, founding researcher at Sakana AI, joins Tim to discuss *Shinka Evolve* — a framework that combines LLMs with evolutionary algorithms to do open-ended program search. The core claim: systems like AlphaEvolve can optimize solutions to fixed problems, but real scientific progress requires co-evolving the problems themselves. GTC is coming, the premier AI conference, great opportunity to learn about AI. NVIDIA and partners will showcase breakthroughs in physical AI, AI factories, agentic AI, and inference, exploring the next wave of AI innovation for developers and researchers. Register for virtual GTC for free, using my link and win NVIDIA DGX Spark (https://nvda.ws/4qQ0LMg) In this episode: • Why AlphaEvolve gets stuck — it needs a human to hand it the right problem. Shinka tries to invent new problems automatically, drawing on ideas from POET, PowerPlay, and MAP-Elites quality-diversity search. • The *architecture* of Shinka: an archive of programs organized as islands, LLMs used as mutation operators, and a UCB bandit that adaptively selects between frontier models (GPT-5, Sonnet 4.5, Gemini) mid-run. The credit-assignment problem across models turns out to be genuinely hard. • Concrete results — state-of-the-art circle packing with dramatically fewer evaluations, second place in an AtCoder competitive programming challenge, evolved load-balancing loss functions for mixture-of-experts models, and agent scaffolds for AIME math benchmarks. • Are these systems act…
Short note on NVIDIA Nemotron 3 Super 120B-A12B, a hybrid Mamba-Transformer MoE model with latent experts and shared-weight MTP.
If Ukraine is the first major drone war, when will there be the first major AI war?
Dive into the realities of AI-assisted coding, the origins of modern fine-tuning, and the cognitive science behind machine learning with fast.ai founder Jeremy Howard. In this episode, we unpack why AI might be turning software engineering into a slot machine and how to maintain true technical intuition in the age of large language models. GTC is coming, the premier AI conference, great opportunity to learn about AI. NVIDIA and partners will showcase breakthroughs in physical AI, AI factories, agentic AI, and inference, exploring the next wave of AI innovation for developers and researchers. Register for virtual GTC for free, using my link and win NVIDIA DGX Spark (https://nvda.ws/4qQ0LMg) Jeremy Howard is a renowned data scientist, researcher, entrepreneur, and educator. As the co-founder of fast.ai, former President of Kaggle, and the creator of ULMFiT, Jeremy has spent decades democratizing deep learning. His pioneering work laid the foundation for modern transfer learning and the pre-training and fine-tuning paradigm that powers today's language models. Key Topics and Main Insights Discussed: - The Origins of ULMFiT and Fine-Tuning - The Vibe Coding Illusion and Software Engineering - Cognitive Science, Friction, and Learning - The Future of Developers RESCRIPT: https://app.rescript.info/public/share/BhX5zP3b0m63srLOQDKBTFTooSzEMh_ARwmDG_h_izk https://app.rescript.info/api/public/sessions/62d06c0336c567d6/pdf Jeremy Howard: https://x.com/jeremyphoward https://www.answer.…
Chip Huyen and I share what we've learned, best practices, and insights at NVIDIA GTC 2025.
The robotics industry stands on the brink of a significant transformation, with many experts – including NVIDIA CEO Jensen Huang – suggesting that we might be approaching a "ChatGPT moment" for robotics. The post Humanoid Robots on the Rise: Industry Advances, Key Players, and Adoption Timelines appeared first on TOPBOTS .
The Yann LeCun et al. (1989) paper Backpropagation Applied to Handwritten Zip Code Recognition is I believe of some historical significance because it is, to my knowledge, the earliest real-world application of a neural net trained end-to-end with backpropagation. Except for the tiny dataset (7291 16x16 grayscale images of digits) and the tiny neural network used (only 1,000 neurons), this paper reads remarkably modern today, 33 years later - it lays out a dataset, describes the neural net architecture, loss function, optimization, and reports the experimental classification error rates over training and test sets. It’s all very recognizable and type checks as a modern deep learning paper, except it is from 33 years ago. So I set out to reproduce the paper 1) for fun, but 2) to use the exercise as a case study on the nature of progress in deep learning. Implementation . I tried to follow the paper as close as possible and re-implemented everything in PyTorch in this karpathy/lecun1989-repro github repo. The original network was implemented in Lisp using the Bottou and LeCun 1988 backpropagation simulator SN (later named Lush). The paper is in french so I can’t super read it, but from the syntax it looks like you can specify neural nets using higher-level API similar to what you’d do in something like PyTorch today. As a quick note on software design, modern libraries have adopted a design that splits into 3 components: 1) a fast (C/CUDA) general Tensor library that implement…
ISSUE: while training Darknet with GE FORCE GTX 1650 using following: CUDA 11.0 cuDNN 8.0.5 OPENCV 4.5 Model starts training with config file details as below for [net] section: [net] # Testing #batch=1 #subdivisions=1 # Training batch=64 subdivisions=16 width=416 height=416 channels=3 momentum=0.949 decay=0.0005 angle=0 saturation = 1.5 exposure = 1.5 hue=.1 learning_rate=0.001 burn_in=1000 max_batches = 6000 policy=steps steps=4800,5400 scales=.1,.1 #cutmix=1 mosaic=1 #:104x104 54:52x52 85:26x26 104:13x13 for 416 When I change the batch from 64 to 32 (reducing it ) coupled with subdivisions increased from 16 to as high as 128, it keeps on going well but then stops after some time as reported above: Error: cuDNN isn't found FWD algo for convolution.