Following are excerpts from the proposal that was submitted to AAAI.
Title: AI for engineering and scientific discoveries
Although AI has been used in scientific discovery for a long time [1,2], the development of large language models and associated generative AI techniques has turbo-charged interest in this topic. A selection of these works [3-10] spans a broad spectrum of scientific and engineering disciplines, including Biology, Physics, Astronomy, and Materials Science and Engineering. However, these works are mostly in the preliminary stage. At the same time, driven by the progress in generative AI, large language models, and AI-assisted scientific computing, there have been several recent funding calls and high level events with respect to the use of AI in engineering and scientific discoveries. These include the Accelerating Computing-Enabled Scientific Discovery (ACED) call by NSF [11], the NSF AI Research Institute calls on AI for Astronomical Sciences and AI for Discovery in Materials Research [12], the DARPA call on Foundation Models for Scientific Discovery (FoundSci) [13], the DARPA call on Scientific Feasibility (SciFy) [14], and the US Army ERDC call on Artificial Intelligence and Machine Learning Models to Inform Materials Discovery. Some of the related high level events are the 2023 AI for scientific discovery workshop organized by the National Academies [15], and the 2024 AAAI Fall Symposium on Integrated Approaches to Computational Scientific Discovery [16].
With the above context, to create and diversify joint efforts towards AI for science and engineering on the emerging frontiers beyond existing efforts, we propose a 2025 AAAI Spring Symposium on AI for engineering and scientific discoveries. The specific goals are (1) to strengthen industry-academia collaborations in materials, manufacturing, and life science subdomains of global interests where domain challenges have not been sufficiently exposed to AI research; and (2) to create discussions and new collaborations in human-machine interaction that develops domain-specific AI-assisted experiences for accelerated knowledge discovery.
Sample areas of interest are as follows:
● How AI is being used and can be used for engineering and scientific discoveries?
● How has the goal of using AI for engineering and scientific discoveries led and can lead to new techniques and developments in AI?
● New understanding of the reducibility of computation on graphs (e.g., for accelerated molecular dynamics) and its use in engineering and scientific discoveries.
● New representation learning to assist discoveries from scientific measurements, e.g., random fields, graphs, and sequences.
● New methods for quantifying uncertainties from AI-assisted scientific discoveries.
● Developing autonomous scientific AI agents for evidence seeking, hypothesis proposing, assessment and experimental cost estimation.
● Ethical consideration of AI-assisted scientific discoveries.
● Out-of-sample and out-of-distribution generation of structured objects.
● Goal-oriented reinforcement learning and generative AI models for targeted structure generation to achieve desired functionalities.
Symposium format:
The two and half days workshop will be composed of keynote talks, oral/paper presentations, poster sessions, and panel discussions. There are two confirmed keynote talks and participants from experienced researchers with diverse backgrounds, including AI and material science program directors from federal funding agencies, senior scientists from pharmaceutical, semiconductor, and manufacturing industries, and leading AI researchers from academia. From the call of papers we will select 10+ contributed talks and posters. Special presentations/panels on funding opportunities and industry needs will be organized to attract participation.
Resolving Submission Conflicts:
This symposium will only accept submissions that explicitly contribute to usages of AI in discovery in materials science and engineering, manufacturing, and life science domains. Submissions that do not explicitly address challenges within these scientific and engineering domains, but rather those related to Agentic AI applications to scientific discovery in general, or related to other domains, should be submitted to the symposium titled "Towards Agentic AI for Science: Hypothesis Generation, Comprehension, Quantification, and Validation.”
Symposium Sponsors: Brewer Science (confirmed), ASU (confirmed), Applied Materials (verbally agreed).
Tentative Program:
Day 1:
9:00 - 9:30 Opening Remarks
9:30 - 10:30 Keynote Speaker 1 (Jared Cobb, US Army ERDC)
10:30 - 11:00 Break
11:00 - 12:30 Contributed Talks 1
12:30 - 2:00 Lunch
2:00 - 3:00 Panel 1
3:30 - 4:00 Break
4:00 - 4:30 Poster Session 1
Day 2:
9:00 - 9:30 Opening Remarks
9:30 - 10:30 Keynote Speaker 2 (TBD)
10:30 - 11:00 Break
11:00 - 12:30 Contributed Talks 2
12:30 - 2:00 Lunch
2:00 - 3:00 Panel 2
3:30 - 4:00 Break
4:00 - 4:30 Poster Session 2
Day 3:
9 - 9:30 Opening Remarks
9:30 - 10:30 Keynote Speaker 3 (Haoda Fu, Amgen)
10:30 - 11:00 Break
11:00 - 12:30 Panel 3
12:30 - 2:00 Lunch
Organizing Committee:
Anton Netchaev, US Army ERDC (Co-chair)
Chitta Baral, Arizona State University (Co-chair)
Lenore Dai, Arizona State University (Co-chair)
Yi Ren, Arizona State University (Co-chair)
Alvaro Velasquez, DARPA
Jianlin Cheng, University of Missouri
Muhao Chen, UC Davis
Prasad Calyam, University of Missouri
Reuben Chacko, Brewer Science
Wu-Sheng Shih, Brewer Science
Yezhou Yang, Arizona State University
...
References:
[1] L. Tari, S. Anwar, S. Liang, J. Cai and C. Baral. Discovering drug-drug interactions: a text mining and reasoning approach based on properties of drug metabolism. Bioinformatics. 26(18):2010. (special issue of ECCB 2010.)
[2] P. Langley. Data-driven discovery of physical laws. Cognitive Science, 5, 31–54. 1981
[3] Pingchuan Ma, Tsun-Hsuan Wang, Minghao Guo, Zhiqing Sun, Joshua B. Tenenbaum, Daniela Rus, Chuang Gan, and Wojciech Matusik. LLM and Simulation as Bilevel Optimizers:
A New Paradigm to Advance Physical Scientific Discovery. https://arxiv.org/pdf/2405.09783.
May 2024
[4] Oskar Wysocki, Magdalena Wysocka, Danilo S. Carvalho, Alex Bogatu, Danilo Gusicuma, Maxime Delmas, Harriet Unsworth, and André Freitas. An LLM-based Knowledge Synthesis and Scientific Reasoning Framework for Biomedical Discovery. https://arxiv.org/pdf/2406.18626. June 2024.
[5] Zechang Sun, Yuan-Sen Ting, Yaobo Liang, Nan Duan, Song Huang, and Zheng Cai.
Knowledge Graph in Astronomical Research with Large Language Models: Quantifying Driving Forces in Interdisciplinary Scientific Discovery. https://arxiv.org/pdf/2406.01391. 2024.
[6] Elahe Khatibi, Mahyar Abbasian, Zhongqi Yang, Iman Azimi, and Amir M. Rahmani. ALCM: Autonomous LLM-Augmented Causal Discovery Framework. https://arxiv.org/pdf/2405.01744. May 2024.
[7] Microsoft Research AI4Science, Microsoft Azure Quantum. The Impact of Large Language Models on Scientific Discovery: a Preliminary Study using GPT-4. December 2023.
[8] Amil Merchant, Simon Batzner, Samuel S. Schoenholz, Muratahan Aykol, Gowoon Cheon & Ekin Dogus Cubuk. Scaling deep learning for materials discovery. Nature volume 624, pages 80–85 (2023)
[9] Shuyi Jia, Chao Zhang, Victor Fung. LLMatDesign: Autonomous Materials Discovery with Large Language Models. https://arxiv.org/pdf/2406.13163 June 2024
[10] Henry W. Sprueill, Carl Edwards, Khushbu Agarwal, Mariefel V. Olarte, Udishnu Sanyal,
Conrad Johnston, Hongbin Liu, Heng Ji, and Sutanay Choudhury. CHEMREASONER: Heuristic Search over a Large Language Model’s. Knowledge Space using Quantum-Chemical Feedback June 2024
[11] NSF call on Accelerating Computing-Enabled Scientific Discovery (ACED).
[12] NSF AI Research Institute call that includes the topics "AI for Astronomical Sciences" and "AI for Discovery in Materials Research" https://new.nsf.gov/funding/opportunities/national-artificial-intelligence-research/nsf23-610/solicitation 2023
[13] DARPA call on Foundation Models for Scientific Discovery (FoundSci). https://www.darpa.mil/program/foundation-models-for-scientific-discovery . 2023.
[14] DARPA call on Scientific Feasibility (SciFy). https://www.darpa.mil/program/scientific-feasibility . 2024.
[15] AI for Scientific Discovery workshop by the US National Scientific Academies. https://www.nationalacademies.org/our-work/ai-for-scientific-discovery-a-workshop 2023.
[16] 2024 AAAI Fall Symposium on Integrated Approaches to Computational Scientific Discovery. http://cogsys.org/symposium/discovery-2024/