LabTech is a free, annual event at Berkeley Lab focused on computing and technology for science and operations. Hosted by the IT Division, LabTech features technical tutorials, interactive discussions, networking opportunities, demonstrations, and more. From collaboration and productivity tools, to high-performance computing and scientific software, everyone will discover a topic of interest!

Abstract: In recent years, the fields of natural language processing and computer vision have been revolutionized by the success of large models pretrained with task-agnostic objectives on massive, diverse datasets (eg. ChatGPT). These so-called ``foundation models'' have enabled transfer learning on entirely new scales, and have outperformed supervised training models across numerous problems.

I will discuss the Polymathic AI initiative, a collaboration between researchers at the Flatiron Institute and scientists around the world. Polymathic AI is designed to spur scientific discovery using similar technology to that powering ChatGPT, we call it "Foundation Models". Using Polymathic AI, scientists will be able to model a broad range of physical systems across different scales. I will present our recent work and soon to be released large scale ML-ready scientific datasets. 

The Berkeley Lab AI for Science Summit (BLASS 24) is an event that brings together AI researchers and industry experts with scientists and national lab staff to explore how AI can drive scientific discoveries. This workshop features presentations and discussions on the latest advancements in AI, with a focus on its applications in science, engineering, and safety. Attendees will have the chance to find out about cutting-edge AI research as applied to real-world scientific and engineering challenges, and discuss the ethical and safety issues related to AI. The summit aims to encourage collaboration across different fields and push forward innovation in AI-driven science.

The workshop will be held in a hybrid format, with livestreamed introductory talks and panel discussions. Breakout sessions will be available exclusively for in-person attendees. Responses to the registration questions will help us understand your focus areas, and based on these answers, the Steering Committee and ALDs will select in-person attendees to ensure diverse perspectives and expertise. 

The first seminar of the semester will be held next Thursday at noon in Cory Hall, Room 373, at UC Berkeley.

We’re thrilled to have Simon Batzner from DeepMind as our speaker - https://bidmap.berkeley.edu/seminars-events .

Don't miss out - be sure to register for the seminar series by emailing bidmap _at_ berkeley.edu with the subject line "Sign Up."

The Department of Energy’s (DOE) Office of Science today announced a new research and development opportunity led by Oak Ridge National Laboratory (ORNL) to advance technologies and drive new capabilities for future supercomputers. This industry research program worth $23 million, called New Frontiers, will initiate partnerships with multiple companies to accelerate the R&D of critical technologies with renewed emphasis on energy efficiency for the next generation of post-exascale computing in the 2029 and beyond time frame.

Funding for the RFP is being provided by the DOE Office of Science’s Advanced Scientific Computing Research Program. The full New Frontiers RFP can be found at https://www.olcf.ornl.gov/newfrontiers/.

This workshop will convene experts in rare and extreme event detection and characterization representing a broad range of application domains and disciplines, including statistics, machine learning, applied mathematics, space weather, materials science, and climate modeling. The goal is for these experts with complementary backgrounds to identify key challenges and opportunities, with an emphasis on methodologies that may be leveraged across domains. The focus on data-driven methods encompasses recent efforts in machine learning, including physics-informed machine learning and generative models, and how such tools may advance rare and extreme event forecasting. The agenda will also include physics-driven approaches, including simulations, both as a source of fundamental insights into the modeling of rare events and as a mechanism for generating data to complement real-world data used to train data-driven models. This two-day workshop will be held at the University of Chicago on November 21-22, 2024. It will feature lectures from experts across the spectrum of disciplines listed above, panel discussions, poster sessions, and lightning talks.

In light of DOE’s recently announced FASST initiative, and anticipating high-level LBNL and DOE workshops later this fall, ESA is holding this workshop to explore the integration of artificial intelligence/machine learning (AI/ML) with chemical and materials science research at LBNL. We aim to identify current scientific challenges, address gaps, and foster the development of innovative AI/ML applications in these fields.The workshop will include presentations from Lab researchers, as well as break-out sessions on focus topics and brainstorming. More information about the workshop can be found on the event website.

The workshop will be held onsite at LBNL. Participants are encouraged to attend in person, though a virtual option will be available. 

The workshop is open to all interested Lab staff, so please feel free to forward this invitation. Please direct questions to esa@lbl.gov. 

Join your colleagues in a new virtual weekly webinar series that will introduce the Berkeley Lab community to AI initiatives, tools, and resources and how the Lab’s research can address national needs. The webinar series kicks off next Thursday,  August 15, 11 a.m to noon, when Jonathan Carter, Associate Lab Director for Computing Sciences, will discuss DOE's new Frontiers in AI for Science, Security, and Technology (FASST) initiative. This event is open to the entire Lab community. A brief registration form is required for organizers to gauge attendance. Future AI webinars in the series will be held every Thursday through Sept. 19. A reminder will be posted in Elements before each webinar.

Synopsis - why FASST?  

The speed and scale with which AI is developing requires investment in a strategic capability now.  Without FASST, the United States stands to lose its competitive scientific edge and ability to maintain our national and economic security, will have a less diverse and competitive innovation AI ecosystem, will not have the independent technical expertise necessary to govern AI, and will lose the nation’s ability to attract and train a talented workforce.  Through FASST, we will meet the mission needs of national security, energy security, and scientific discovery that will support sustained economic prosperity for the nation for decades to come.

DOE’s Enabling Infrastructure  

DOE and its 17 national laboratories are uniquely positioned to develop AI capabilities for the nation, leveraging key enabling components:  Data, Computing, Workforce, Partnerships.   

FASST initiative leverages DOE’s enabling infrastructure to deliver key assets for the national interest: Advance National Security, Attract and build a talented workforce, Harness AI for Scientific Discovery, Address Energy Challenges, Develop technical expertise necessary for AI governance.

DOE Roadmap announcement (July 16 2024): LINK

Sparsitute brings leading researchers across the nation working on various aspects of sparsity together to accelerate progress and impact. Our ambitious research agenda will drastically advance the state-of-the-art in sparse computations both as a unified topic and within three broad pillars: sparse and structured matrix computations, sparse tensor problems, and sparse network problems, as well as their interconnections.

This provides new capabilities for problems that already have sparsity acknowledged but not fully harnessed and problems where sparsity has not been discovered. Our institute will play a pivotal role in analyzing the data at system, micro, and nano scales captured from different earth-shot applications. Thus, it will have a synergistic impact on scientific computing and scientific data analysis investments by DOE.

There may be opportunities for collaboration and funding through this center and we (NSD) think the path forward is to hold a small workshop with CSA in the coming months - please contact nsd-advanced-computing-wg@lbl.gov if you have ideas, questions etc.

The workshop program will cover a broad range of current software and computing projects in the Nuclear Physics domain. Discussions will include general issues such as the role of research vs. infrastructure software, development of common software ecosystems, software sustainability best practices, and workforce development. Consideration of new opportunities will be carried out in the context of existing and currently planned DOE and NSF programs that support advanced NP computing and cyber-infrastructure. 

HPDF: A New Element in the ASCR Ecosystem

Leading-edge science demands integration, and the U.S. Department of Energy’s science programs increasingly require advanced computation, data solutions, and networking. The High Performance Data Facility, a first-of-its-kind project, is envisioned as a state-of-the-art resource for data science and research. HPDF will join other Advanced Scientific Computing Research facilities engaged in implementing the DOE’s Integrated Research Initiative.

As the cornerstone of the IRI, HPDF is committed to supporting scientific research patterns and data stewardship within the nation’s research communities. Along with the other ASCR facilities – Energy Sciences Network, Argonne Leadership Computing Facility, Oak Ridge Leadership Computing Facility, and the National Energy Research Scientific Computing Center – HPDF will help forge the IRI’s foundational infrastructure.

The project’s hub-and-spoke model will maximize availability, resilience, and accessibility. The Hub core infrastructure, physically located at Thomas Jefferson National Accelerator Facility and Lawrence Berkeley National Laboratory, will support centralized resources. Multiple mission-application Spokes will connect to the Hub via ESnet.

The integrated ASCR ecosystem, with HPDF at its heart, will strengthen the nation’s scientific enterprise and support a network of global scientific innovation.

See HPDF: Status and Plans - June 2024