Generative AI are revolutionary technologies impacting our daily human-computing interactions creating new content that matches human creativity. Among these, Large language models (LLM), in particular OpenAI’s generative Generative Pre-trained Transformer (GPT) foundation models and Google's Bidirectional Encoder Representations from Transformers (BERT), have become a ubiquitous topic in the present era. Therefore, there is a strong need to understand their impact, limitations, responsible use, and broader implications for Oak Ridge National Laboratory's (ORNL) scientific mission.
Aims and Scope
ORNL's Generative AI for Science Workshop series, invites the scientific community to share current challenges, requirements and opportunities for the ethical use of generative AI technologies in our mission. Our goal is to provide a venue to educate and exchange research and development ideas, collaborations and investments around the current state-of-the-art in these relatively new technologies. We welcome lightning talk proposals and panel participation from the wide range of experimental, observational, high-performance computing (HPC) projects at ORNL. We will summarize our ideas, findings and key opportunities in a subsequent report that we will share with the community and interested participants.
Registration
While this is a hybrid event, we encourage in-person participation for the tutorial, working lunch session, lightning talks, panel discussions to learn more about Generative AI and build a community at ORNL.
Registration Link: Please register by July 17th filling out this form. Registration is only opened at this point for ORNL employees (with a @ornl.gov either foreign national or US-citizens) and EXTERNAL US-Citizens for virtual participation. In person participation is closed.
Previously registered External (virtual or in-person) participants need to apply for a Personal Access System (PAS). Participation is subject to approval. Contact the organizers for questions.
The registration form includes the option to propose a lightning talk focusing on the requirements for scientific areas that are representative of ORNL.
- Registration is required for in-person and virtual meeting option via Zoom
- We encourage early registration for in-person participation due to the venue’s limited capacity (closed)
Call for Lightning Talks
We invite ORNL participants to present a 3 to 5 minute lightning talk that aligns with the scope of the workshop (option provided in the registration form).
Topics:
- Scientific application requirements for the use of Generative AI
- Responsible use of generative AI for obsevational, experimental and computational science
- Training and educational challenges and requirements
- Ethical and legal implications on the use of Generative AI
- Guardrails for generative AI models
- Large language models (LLMs): GPT, Bard, etc.
- Impact of LLMs in the scientific discovery process
Agenda
| Time | Session | Presenter |
|---|---|---|
| Morning | ||
| 10:00am-10:20am | Opening Remarks | Susan Hubbard, Deputy for Science and Technology, ORNL |
| 10:20am-10:35am | GitHub Copilot for HPC programming | William Godoy |
| 10:35am-10:55am | Considerations for applying large language models to clinical text | John Gounley |
| 11:00am-11:45am | Lightning Talks and Panel Session I: Generative AI and LLMs | Moderator: Andrea Delgado |
| Generative AI in System Safety Engineering | Kelly Mahoney | |
| Document Information Retrieval using LLMs | Sudarshan Srinivasan | |
| LLM for Science: Leveraging NLP and IE on Scientific Publications for Knowledge Discovery | Tirthankar Ghosal | |
| Developing deep generative models for scientific data | Jong Choi | |
| Mona Lisa’s smile or the French Riviera? | Suhas Sreehari | |
| 11:45am-12:00pm | Break | |
| Lunch Talks | Session Chair: Jeffrey Vetter, ASCR Section Head and Corporate Fellow, ORNL | |
| 12:00pm-12:20pm | Keynote: “Securing the Future of AI: Understanding and Mitigating Threats to AI Data, Models, and Processes” | Prasanna Balaprakash, Director of AI Programs, ORNL |
| 12:20pm-12:45pm | Invited Talk: “Exploring the Use of Agents in Chemistry” | Samantha Cox, University of Rochester |
| 12:45pm-1:00pm | Break | |
| Afternoon | ||
| 1:00pm-2:30pm | Lightning Talks and Panel Session II: Scientific Applications | Moderator: Pedro Valero-Lara |
| Exploring Foundation Model for Climate Applications | Valentine Anantharaj | |
| GANus Pauling: a generative model for protein structure design | Julie Mitchell | |
| Scientific Application Requirements for the Use of Generative AI in Atomistic Materials Modeling | Massimiliano (Max) Lupo Pasini | |
| Toward Multimodal Foundation Models for GeoAI | Philipe Ambrozio Dias | |
| Advancing Molecular Optimization: A Generalized Strategy for Exploring Chemical Space using Generative Machine Learning Models | Debsindhu Bhowmik | |
| Inorganic Material Design Using Generative Adversarial Networks | Kadir Amasyali | |
| Inverse design of molecular structure for target optical properties using generative AI | Pilsun Yoo | |
| 2:30pm-2:45pm | Break | |
| 2:45pm-3:55pm | Lightning Talks and Panel Session III: Foundational Models | Moderator: Oscar Hernandez |
| Pretraining Large Language Models at OLCF | Junqi Yin | |
| Verification of Generative AI – can formalism help? | Keita Teranishi | |
| Blackout Diffusion: Diffusion Models in Discrete State Spaces | Zach Fox | |
| Generative AI based run-time settings of tiled MatRIS algorithms | Narasinga Rao Miniskar | |
| Improving the autoencoder interpolation via dynamic optimal transport | Xue Feng | |
| Leveraging Constrained Generative Adversarial Networks (GANs) for 3D Image Reconstruction and Segmentation in Scientific Imaging | Amir Ziabari | |
| Differentially private language model training | Chris Stanley | |
| 3:55pm-4:00pm | Closing Remarks, Adjourn |
Organizers
- Prasanna Balaprakash
- Edmon Begoli
- Andrea Delgado
- William Godoy
- Oscar Hernandez
- Dalton Lunga
- Thomas Potok
- Juan Restrepo
- Amir Sadovnik
- Keita Teranishi
- Pedro Valero-Lara
- Jeffrey Vetter
- Donna Wilkerson