Henrique Malvar

The 7th workshop on Neural Scaling Laws is about to start: https://lnkd.in/gQSB5k8v Scaling laws, emergency, alignment, transfer learning, with a focused track on multilingual models. Looking forward to kick it off with Irina Rish and discuss all these topics with Yuhai Tu, Paul Bogdan, Hailey Schoelkopf, Darshil Doshi, Rio Yokota, Neha Sengupta, Preslav Nakov, Tatiana Shavrina, PhD, Yishi Xu. Join in person or virtually! #icml2024 #ai

37

A number of folks asked me about getting the slides that John Fragalla, M.E. presented at the SC24 DDN Data Summit two weeks ago on "NCP Solutions: An Optimized Cloud Data Architecture with DDN" So I'll overdeliver - here's the FULL video of John's presentation! The whole video is worth watching - but I'll call out a few moments you should review, even if you're short on time: - 8:32 - The I/O story - which is CRITICAL to how your AI will operate - 11:26 - Checkpointing and Model Load - also important with keeping AI running - 16:34 - How architecting for peak performance saves MILLIONS John is a great friend, and if you'd like to dig deeper into this, drop me a line at [email protected] - I'll bring John along. :) #ArtificialIntelligence #AI #NCP #Cloud DDN NVIDIA

42

1 Comment

A very impressive set of papers from Hewlett Packard Labs on sustainability, renewable energy and generative AI topics to be presented at this week’s NeurIPS conference. Great to see further work on spatiotemporal carbon optimisation, reinforcement learning for data centre sustainability, advancing research on nuclear fusion and expanding research with Carnegie Clean Energy. More details, links to papers and videos in this blog post.

4

⚠️ 𝗔 𝗖𝗿𝗶𝘁𝗶𝗰𝗮𝗹 𝗟𝗼𝗼𝗸 𝗮𝘁 𝗨.𝗦. 𝗘𝘅𝗽𝗼𝗿𝘁 𝗖𝗼𝗻𝘁𝗿𝗼𝗹𝘀: The latest paper from Ritwik Gupta & team at the Berkeley Risk and Security Lab and the Berkeley Artificial Intelligence Research (BAIR) Lab, “Whack-a-Chip: The Futility of Hardware-Centric Export Controls,” highlights a troubling reality—current U.S. strategies to limit China’s AI advancements through semiconductor export controls are falling short. Key findings: * 𝗘𝘃𝗮𝘀𝗶𝘃𝗲 𝗧𝗮𝗰𝘁𝗶𝗰𝘀: Despite restrictions, Chinese firms like Tencent are accessing high-end chips, using innovative approaches to train state-of-the-art models like Hunyuan-Large on non-export-controlled hardware, bypassing GPU RDMA soft locks in NVIDIA’s consumer-grade cards with custom drivers, and even gaining direct access to export-controlled cards. * 𝗘𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝗰𝘆 𝗚𝗮𝗶𝗻𝘀 𝘁𝗵𝗲 𝗘𝗱𝗴𝗲: Advances in machine learning efficiency—such as mixed-precision training, mixtures of experts, quantization, and model & data parallelism—are driving breakthroughs and reducing dependence on cutting-edge hardware. Notably, many of these techniques originate from GPU-starved researchers outside of China. * 𝗣𝗼𝗹𝗶𝗰𝘆 𝗜𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀: The paper challenges the efficacy of existing controls and emphasizes the need for more nuanced strategies to maintain technological leadership in the AI arms race. This is a sobering read but an essential one for U.S. policymakers and industry leaders navigating the rapidly shifting AI landscape. 💡 Read the full paper here: https://lnkd.in/gAFQj5yV

19

If you kind of said 15 years for very useful quantum computers, that would probably be on the early side. If you said 30, it’s probably on the late side,” Huang said in a question-and-answer session during Nvidia’s analyst day. “If you picked 20, I think a whole bunch of us would believe it. The basic problem of input and output was not solved to the best of my knowledge. When it will be solved, a new exciting era will start https://lnkd.in/d7vnnGVB #research #quantumcomputing #AI

3

3 Comments

Polaris 2.0 from Hippocratic AI is here. Safer than ever. 3T+ parameter constellation architecture with voice latency equal to Polaris 1.0 at 1T+ parameters. Read a summary of the changes from our chief scientist Subhabrata (Subho) Mukherjee's blog: https://lnkd.in/gXweaWe5

279

1 Comment

The new release of the Alluxio Enterprise AI data orchestration platform makes it easier to use GPU-based systems for training and operating AI applications and to provision AI/ML systems with data at HPC levels. https://bit.ly/3LoLTkX

1

Arm is upstreaming #Kleidi technology into #PyTorch with amazing effect - up to 3x faster CPU #inference in Eager mode, and up to 2x in Graph mode. These optimizations make running #LLMs on CPU more practical than ever. Check out our blog for more details! #Arm #Neoverse #Graviton https://lnkd.in/gvrh67v2

17

Emerging technologies don’t simply materialize fully formed; they require the establishment of supply chains, the development of manufacturing processes, and the evolution of the technology itself. Along this journey, the cost incurred includes learning from the ten unsuccessful paths before arriving at the eleventh one that proves effective…a perceptive read if you are working in the Spatial Computing space

20

On Thursday, 6/13 at 9 AM PT, learn about the game-changing benefits of #AI quantization. Qualcomm AI Research experts Mart van Baalen and Abhijit Khobare will share how state-of-the-art LLM quantization techniques like vector quantization address the challenges of implementing on-device AI efficiently and accurately. https://bit.ly/3KvJzs1

75

1 Comment

Join Carnegie Mellon University's Philip Koopman on Sept 11, 10 AM EDT, for insights on leveraging #AI effectively. Learn about: • Classification & generative outputs • Addressing bias & safety • The 90/10 principle • Real-world Machine Learning applications Understand AI's potential and limitations. Register now! https://lnkd.in/gvhVk33y #MachineLearning #TechInnovation #BOSS2024 #ArtificialIntelligence #DeepLearning #Semiconductor #ML

8

For his debut TDS article, Jeremi Nuer presents a thoughtful walkthrough of a recent paper on model interpretability: "We want to have finer-tuned control over LLMs. Being able to detect when certain features are active, and tune how active they are in the middle of generating output, is an amazing tool to have in our toolbox."

18

My recent blog post with Iris Fang on how NeuroPace ML workloads is leveraging Google Cloud is now live.

31

1 Comment

Training computer vision models depends on quantifying images’ similarity, but similarity can be a matter of concepts — not just pixel values. At CVPR 2024, Amazon researchers showed how to leverage vision-language models to quantify images’ conceptual similarity according to the length of the descriptions required to discriminate them: images easily discriminated by short descriptions are not very similar, but images that require a lot of text to distinguish them are similar. #ComputerVision #ExplainableAI

50

1 Comment

Prem Natarajan, PhD and Joey C. in how Capital One partner with NVIDIA and AWS to accelerate GenAI use cases: "Our partnership with NVIDIA focuses on leveraging the right software to accelerate Generative AI use cases at the bank." Awesome work by Anthony Murphy Annie Prasanna Surla Jiahong Liu Timothy Ma James Allen

61

🚀 NVIDIA NIM on Bedrock Marketplace 🚀 We’re thrilled to announce that NVIDIA NIM are now available on Amazon Bedrock Marketplace, SageMaker JumpStart, and AWS Marketplace. With this expansion, developers can now deploy high-performance, NVIDIA-optimized inference solutions for generative AI models across key AWS services, delivering: ✅ Faster AI inference ✅ Lower latency ✅ Cost efficiency From Meta’s Llama 3 to NVIDIA Nemotron and more, over 100 prebuilt NIM microservices are now at your fingertips. Tailored for scalability and secure deployment, NIM brings unmatched performance to models running on EC2, EKS, SageMaker, and Amazon Bedrock. This collaboration between NVIDIA and AWS accelerates innovation across industries, from healthcare to manufacturing, ensuring enterprises can meet the growing demand for generative AI applications. 💡 Ready to unlock the power of NVIDIA NIM on AWS? Explore the catalog, request a developer license, or start a 90-day NVIDIA AI Enterprise trial to experience the difference firsthand. Another great project by this wonderful team! Abhishek Sawarkar Jiahong Liu Kshitiz Gupta Timothy Ma Saurabh Trikande Charlie Huang Rohil Bhargava Let’s power the next wave of AI together! #GenerativeAI #NVIDIA #AWS #Bedrock #AIInference #AWSreInvent2024

44

1 Comment