Source URL: https://cloud.google.com/blog/topics/public-sector/google-public-sector-supports-ai-optimized-hpc-infrastructure-for-researchers-at-caltech/
Source: Cloud Blog
Title: Google Public Sector supports AI-optimized HPC infrastructure for researchers at Caltech
Feedly Summary: For decades, institutions like Caltech, have been at the forefront of large-scale artificial intelligence (AI) research. As high-performance computing (HPC) clusters continue to evolve, researchers across disciplines have been increasingly equipped to process massive datasets, run complex simulations, and train generative models with billions of parameters. With this powerful combination, researchers have accelerated scientific discovery across diverse use cases including genomic analysis, drug discovery, weather forecasting, and beyond.Modern research workloads, driven by AI and HPC, demand processing of structured and unstructured data at an unprecedented scale, while maintaining sub-millisecond storage latency, enterprise-level security and compliance, and reproducibility despite varying hardware and software configurations. This presents significant technical challenges for both researchers and supporting departments.To accelerate scientific discovery in the AI era, Google Public Sector has announced it will support AI-optimized HPC infrastructure for researchers at Caltech. This new initiative furthers Caltech’s mission to expand human knowledge and benefit society through research integrated with education.This support provides Caltech researchers four key resources:A workhorse of diverse processor types, including Graphics Processing Units (GPUs) and Google’s custom-design Arm-based processors (Axion) and Cloud Tensor Processing Units (TPUs) for AI acceleration and intense workloads.Access to Google’s first party datasets like AlphaFold, Earth Engine, Google Maps Platform and VirusTotal to accelerate discoveries across all disciplines.A fully-managed, unified AI development platform, Vertex AI Platform, which includes Vertex AI Agent Builder and 200+ first-party (Gemini, Imagen 3), third-party, and open (Gemma) foundation models in Model Garden.Dedicated campus training and workshops for students, researchers, and supporting teams, enabling them to increase AI literacy and adoption.Google Public Sector and Caltech will integrate this AI-optimized infrastructure with Caltech’s existing HPC research environments to provide researchers instant access while maintaining their existing data and workloads.One of the first initiatives that will be powered by this new AI infrastructure will be led by Dr. Babak Hassibi, Mose and Lillian S. Bohn Professor of Electrical Engineering and Computing and Mathematical Sciences at Caltech. Dr. Hassibi’s research focuses on making AI models more efficient, which is crucial for the advancement of the field. Current large language models (LLMs) can have billions or trillions of parameters. In efforts to make models even more efficient and useful. Dr. Hassibi’s proposal uses Vertex AI and has the potential to make AI more accessible and sustainable.”We will be using Vertex AI to develop training methods on TPUs that incorporate pruning, quantization, and distillation, as well as considerations regarding resilience to attacks, into the training process itself. The former has the potential to significantly reduce inference time costs of the trained models, making AI much more accessible and sustainable. The latter can markedly improve the safety of systems that employ AI. Both will allow AI to move to the edge. In addition to the practical benefits, the work will inform the theoretical studies of AI models, in particular, their generalization performance and the limits of their compressibility, which is a major focus of my research group," said Dr. Babak Hassibi.By providing access to advanced AI and planet-scale infrastructure, this support enables Caltech researchers to continue to push scientific boundaries, investigate complex problems, and develop innovative solutions.“We are living a time when we need to answer bigger questions faster, and do more with less. Google Public Sector is excited to support Caltech to build an AI-optimized infrastructure that will lead scientific discoveries across all domains for the best of all our constituents,” said Reymund Dumlao, Director of State & Local Government and Education at Google Public Sector.At Google Public Sector, we’re passionate about applying the latest cloud, AI and security innovations to help you meet your mission. Subscribe to our Google Public Sector Newsletter to stay informed and stay ahead with the latest updates, announcements, events and more.For more information about Caltech’s research heritage and centers, visit https://www.caltech.edu/research
AI Summary and Description: Yes
Summary: This text discusses a collaborative initiative between Google Public Sector and Caltech to provide AI-optimized high-performance computing (HPC) infrastructure aimed at accelerating scientific research. The partnership emphasizes the integration of advanced AI resources and security for processing large-scale data, which is relevant for professionals concerned with cloud computing, AI development, and infrastructure security.
Detailed Description:
This text highlights a significant collaboration focused on enhancing AI research capabilities within high-performance computing (HPC) environments. Here are the key points:
– **Partnership Overview**: Google Public Sector and Caltech have joined forces to create AI-optimized HPC infrastructure, facilitating improved access and capabilities for researchers.
– **Resource Provision**: Caltech researchers will benefit from:
– A diverse range of processor types (GPUs, Google’s Arm-based processors, TPUs) designed for AI workloads.
– Access to Google’s first-party datasets like AlphaFold, Earth Engine, etc., crucial for various scientific investigations.
– A fully-managed AI development platform (Vertex AI Platform), comprising multiple advanced models.
– Training workshops aimed at increasing AI literacy among users.
– **Focused Research**: Dr. Babak Hassibi’s research will utilize this infrastructure to develop more efficient AI models, tackling the complexity associated with LLMs. His approach includes:
– **Training methods** integrating pruning, quantization, and distillation to optimize models for accessibility and sustainability.
– Enhancements in security through resilience against potential attacks during model training.
– **Impact on Scientific Discovery**: The initiative is expected to allow researchers at Caltech to:
– Push the boundaries of scientific investigation efficiently, using less resource-intensive methods.
– Investigate complex scientific problems while ensuring data integrity and security through enterprise-level compliance.
– **Practical Implications**: This effort epitomizes the blending of cutting-edge cloud and AI technologies to advance industries and disciplines. The integration aims to streamline research processes while promoting safer AI practices.
– **Future Outlook**: The continued collaboration between Google and Caltech positions them at the forefront of AI research, with implications for scalability and sustainability in scientific studies.
This initiative underscores the importance of combining advanced computing resources with security considerations, thus offering insight into best practices for AI and cloud infrastructure development while meeting prevailing compliance and data governance standards.