CONVOHPC 2023

1:30 - 1:40 PM

Opening Remarks

Hari Subramoni and Aamir Shafi The Ohio State University

1:40 - 2:10 PM

Speaker: Brandon Biggs, Idaho National Laboratory

Session Chair: Hari Subramoni, The Ohio State University

Title: Integrating Conversational Artificial Intelligence Systems into Science Gateways

Abstract: Science gateways are altering the manner in which people interact with high performance computing (HPC) by providing a web browser based interface to advanced computing platforms. In particular, science gateways lower the barrier to using HPC by simplifying the process of submitting workloads to such systems and by offloading the efforts required to use HPC to the maintainers of the system. While science gateways decrease the time-to-science that comes with using such advanced systems, progress can still be made in improving the user's experience. Large language models can be utilized to support staff and assist users. A certain class of user inquiries can be supplemented or even resolved by these language models. This talk will explore those inquiries via integrated conversational artificial intelligence systems commonly found in non-HPC service workflows.

Speaker Bio: Brandon Biggs is a High Performance Computing (HPC) Systems Administrator in Idaho National Laboratory’s Advanced Scientific Computing Department, part of the Nuclear Science & Technology directorate. He has experience with HPC system management, software integration, full stack development, and machine learning. Before joining INL in 2019 he was a systems administrator for the College of Science and Engineering at Idaho State University, managing scientific data and computational systems, as well as educational computing resources for students. He also worked as a Graduate Research Assistant and Computer Science Outreach Coordinator while at Idaho State. He completed his master’s degree in 2022 and his B.S in 2018, both in computer science from Idaho State University.

2:10 - 2:40 PM

Speaker: Beth Plale, Indiana University

Session Chair: Hari Subramoni, The Ohio State University

Title: Trustworthiness in Conversational Interfaces for the Edge-Cloud-HPC Continuum

Abstract: Issues of bias, inequity, and manipulation have brought AI ethics into wide concern. Ethical issues may also manifest from the cyberinfrastructure for AI. This is because infrastructure is embedded in other human structures and social arrangements, and in other technologies. Social order relies on infrastructure which is largely invisible until breakdown. Cyberinfrastructure for AI must meet the critical ethical challenges of supporting human endeavors using methods that minimize the embedding of AI bias and inequity.

In this talk I will consider trustworthiness issues in conversational interfaces for HPC. These conversational interfaces may use AI generative models and certainly depend on representations of knowledge in the form of knowledge graphs and ontologies. Sources will matter. Transparency will matter. Our work on the realization of AI transparency in infrastructure is through AI Model Cards and is grounded in the US National Science Foundation (NSF) funded Intelligent Cyberinfrastructure with Computational Learning in the Environment (ICICLE) AI Institute. ICICLE innovates in use inspired science, placing at-edge concerns at the center of cyberinfrastructure development. Our complementary focus on an aware workforce and a transparent infrastructure are critical cornerstones our newly developed framework for ethical AI.

Speaker Bio: Dr. Beth Plale, the Michael A and Laurie Burns Professor of Computer Engineering, is Department Chair of the Department of Intelligent Systems Engineering at Indiana University Bloomington (IU). Plale additionally serves as the Executive Director of the Pervasive Technology Institute and Founding Director of the Data To Insight Center. Plale’s interests computational and data infrastructure, open science, provenance & reproducibility, AI ethics, and data accountability. Plale served at the US National Science Foundation (NSF) in a policy position working on open science (2017 -2020). Dr. Plale’s postdoctoral studies were at the Georgia Institute of Technology and her PhD is in computer science from the Watson School of Engineering at the State University of New York Binghamton.

Plale is a founder of the Indiana University Center of Excellence for Women & Technology (CEWIT) and a founding director of the Hathi Trust Research Center (HTRC). She is an original founder of the international Research Data Alliance (RDA), served as inaugural chair of the RDA Technical Advisory Board (TAB), and lead of RDA-US efforts. Plale received the Early Career award from the Department of Energy (DOE) and is a senior member of ACM and IEEE.

2:40 - 3:00 PM

Speaker: Pouya Kousha, The Ohio State University

Session Chair: Hari Subramoni, The Ohio State University

Title: Live Demo: Conversational AI Interface for HPC Workloads

Abstract: This talk will introduce SAI, a Speech Assistant Interface for High-Performance Computing (HPC) systems. Integrated with OnDemand, SAI allows real-time execution of HPC MPI benchmarks on a cluster. Its main features include job script generation, job execution, and the ability to run jobs across diverse architectures. SAI auto-detects package dependencies and supports package installation. It also provides default values for job variables, checks for completeness, and verifies arguments, reducing errors. The natural language interface adapts to the user's HPC needs, promoting increased system usage through its user-friendly design.

Speaker Bio: Pouya Kousha is a PhD candidate at The Ohio State University, USA, contributing to the Nowlab research group led by Prof. DK. Panda and Dr. H. Subramoni. His research targets real-time analysis, monitoring, and profiling of High-Performance Computing (HPC) systems, with his thesis centered on developing AI-enabled tools for HPC communication analysis. Pouya is the lead developer for OSU INAM, a tool for HPC cluster monitoring downloaded more than 6k times and deployed at various HPC centers globally. With interests in utilizing AI for HPC, distributed systems, and profiling tools, he has reviewed for over 20 HPC conferences and have been a program committee member of various HPC conferences.

3:00 - 3:30 PM

Break

3:30 - 4:00 PM

Speaker: Matthew Lange, IC-FOODS

Session Chair: Hari Subramoni, The Ohio State University

Title: Talk reality to me: Grounding Chat in evolving real world facts

Abstract: The successful launch of ChatGPT broadened interest in Chat as a technology interface, and transformed both possibilities and expectations for how we interact with technology writ large. Yet real-world biases and statistical artifact-driven hallucinations create untrustworthiness in Chat technologies, bringing concern to the large language models (LLMs) underpinning their overall success as a new mode of human computer interaction. Domain scientists wanting to leverage AI in general and Chat in particular, need up-to-date information and verifiable suggestions that are grounded in reality, consistent with real-world knowledge, and drawn across to ever-expanding and evolving knowledge bases of scientific facts. Ontology based knowledge graphs form the bases for AI cyberinfrastructures leverageable across domain endeavors. Considered in this talk are examples of underlying domain ontologies and semantic infrastructure components which, when synergized with LLMs, form architectures for advancing Chat technologies across HPC for use by AI-enabled domain scientists. Chat technologies, together with pluggable, reusable, and intuitive visual interfaces, are part of the US National Science Foundation (NSF) funded Intelligent Cyberinfrastructure with Computational Learning in the Environment (ICICLE) AI Institute. At ICICLE, we are democratizing AI beginning with domain science use cases. One of the ways we are democratizing AI is by lowering technical barriers to entry for domain scientists who need to be neither AI nor HPC experts in order to leverage these computational resources in the advancements of science.

Speaker Bio: Dr. Matthew Lange is the founder, CEO, and Chief Science Officer of the International Center for Food Ontology Operability Data and Semantics (IC-FOODS) a 501c3 non-profit research institute. IC-FOODS was established with a dual mission: first, to build an Internet of Food (IoF) and a Semantic Web of Food (SWoF) that enable greater transparency and traceability across food systems; and second, to leverage these infrastructures in the democratization of food systems by advance science and improve knowledge of food across the ag < > food processing < > distribution < > diet < > health knowledge continuum.

4:00 - 4:30 PM

Speaker: Hari Subramoni, The Ohio State University

Session Chair: Hari Subramoni, The Ohio State University

Title: SAI: AI-Enabled Speech Assistant Interface for Science Gateways in HPC

Abstract: High-Performance Computing (HPC) is increasingly being used in traditional scientific domains as well as emerging areas like Deep Learning (DL). This has led to a diverse set of professionals who interact with state-of-the-art HPC systems. The deployment of Science Gateways for HPC systems like Open On-Demand has a significant positive impact on these users in migrating their workflows to HPC systems. Although computing capabilities are ubiquitously available (as on-premises or in the cloud HPC infrastructure), significant effort and expertise are required to use them effectively. This is particularly challenging for domain scientists and other users whose primary expertise lies outside of computer science. In this paper, we seek to minimize the steep learning curve and associated complexities of using state-of-the-art high-performance systems by creating SAI: an AI-Enabled Speech Assistant Interface for Science Gateways in High Performance Computing. We use state-of-the-art AI models for speech and text and fine-tune them for the HPC arena by retraining them on a new HPC dataset we create. We use ontologies and knowledge graphs to capture the complex relationships between various components of the HPC ecosystem. We finally show how one can integrate and deploy SAI in Open OnDemand and evaluate its functionality and performance on real HPC systems. To the best of our knowledge, this is the first effort aimed at designing and developing an AI-powered speech-assisted interface for science gateways in HPC.

Speaker Bio: Dr. Hari Subramoni is an assistant professor in the Department of Computer Science and Engineering at the Ohio State University. His current research interests include high performance interconnects and protocols, parallel computer architecture, network-based computing, exascale computing, network topology aware computing, QoS, power-aware LAN-WAN communication, fault tolerance, virtualization, big data, deep learning and cloud computing. He has published over 100 papers in international journals and conferences related to these research areas. He has been actively involved in various professional activities in academic journals and conferences. Dr. Subramoni is doing research on the design and development of MVAPICH2 (High Performance MPI over InfiniBand, iWARP and RoCE) and MVAPICH2-X (Hybrid MPI and PGAS (OpenSHMEM, UPC and CAF)) software packages. He is a member of IEEE & ACM.

Closing Remarks

Hari Subramoni and Aamir Shafi The Ohio State University