Uploaded October 2026 | Updated November 2026, 4 days ago
Models, Inference, and Algorithms | September 30, 2026
Broad Institute of MIT and Harvard
Primer: Agentic AI for Scientific Discovery (Part 1)
George Karniadakis, PhD
Professor of Applied Mathematics and Engineering, Brown University
Seminar: Agentic AI for Scientific Discovery (Part 2)
Nazanin Ahmadi
PhD candidate in Biomedical Engineering, Brown University
Abstract:
Scientific creativity rarely emerges from isolated computation; it often arises through interactions among ideas, models, data, and researchers. While current agentic AI systems can automate increasingly complex scientific workflows, a more fundamental challenge remains: Can AI systems generate genuinely novel hypotheses and contribute to creative scientific discovery? In this talk, we explore agentic AI through the lens of Scientific Machine Learning (SciML), where data-driven learning is coupled with mathematical structure and physical knowledge. We discuss how physics-informed neural networks, neural operators, and related SciML frameworks can serve as computational engines for scientific reasoning and model discovery. We then present an agentic framework designed to support iterative scientific discovery through memory, information sharing, multi-agent reasoning, hypothesis generation, mathematical and computational validation, and self-improvement. Rather than relying on a single model or agent, the system enables multiple specialized agents to propose, critique, test, and refine candidate scientific explanations. Through examples ranging from gene-expression trajectories to cancer-treatment dynamics, we demonstrate how these systems can integrate data, prior knowledge, and mathematical structure to uncover interpretable models and generate new scientific hypotheses. Ultimately, we ask how Agentic AI and Scientific Machine Learning can move beyond automating existing scientific tasks toward identifying unexplained patterns, formulating testable hypotheses, and proposing new directions for scientific research.
Biographies:
George Karniadakis is an elected member of the National Academy of Engineering, National Academy of Arts and Sciences, and a Vannevar Bush Faculty Fellow. He received his S.M. and Ph.D. from Massachusetts Institute of Technology (1984/87). He was appointed Lecturer in the Department of Mechanical Engineering at MIT and subsequently he joined the Center for Turbulence Research at Stanford / Nasa Ames. He joined Princeton University as Assistant Professor in the Department of Mechanical and Aerospace Engineering and as Associate Faculty in the Program of Applied and Computational Mathematics. He was a Visiting Professor at Caltech in 1993 in the Aeronautics Department and joined Brown University as Associate Professor of Applied Mathematics in the Center for Fluid Mechanics in 1994. After becoming a full professor in 1996, he continued to be a Visiting Professor and Senior Lecturer of Ocean/Mechanical Engineering at MIT. He is an AAAS Fellow (2018-), Fellow of the Society for Industrial and Applied Mathematics (SIAM, 2010-), Fellow of the American Physical Society (APS, 2004-), Fellow of the American Society of Mechanical Engineers (ASME, 2003-) and Associate Fellow of the American Institute of Aeronautics and Astronautics (AIAA, 2006-). He received the William Benter Prize (2026), SES G.I. Taylor medal (2014), the SIAM/ACM Prize on Computational Science & Engineering (2021), the Alexander von Humboldt award in 2017, the SIAM Ralf E Kleinman award (2015), the J. Tinsley Oden Medal (2013), and the CFD award (2007) by the US Association in Computational Mechanics. His h-index is 171 (highest in Applied Mathematics) and he has been cited over 184,000 times. In 2025 he was named by Research.com the number one math scientist in the world. (Speaking from 9 - 9:45 a.m.)
Nazanin Ahmadi is a PhD candidate in Biomedical Engineering at Brown University. Her current work focuses on agentic AI in scientific machine learning settings, with applications in systems biology, systems pharmacology, and pharmacometric modeling.
About MIA:
The Models, Inference & Algorithms (MIA) Initiative at the Broad Institute supports learning and collaboration across the interface of biology and medicine with mathematics, statistics, machine learning, and computer science. Our weekly meetings are open and pedagogical, emphasizing lucid exposition of computational ideas over rapid-fire communication of results.
MIA is hosted by the Eric and Wendy Schmidt Center at the Broad Institute.
Relevant Links:
MIA Website: broadinstitute.org/mia
MIA YouTube Playlist: broad.io/MIAPlaylist
Brown University CRUNCH Group: https://www.brown.edu/research/projects/crunch/
Copyright Broad Institute, 2026. All rights reserved.
Models, Inference, and Algorithms | September 30, 2026
Broad Institute of MIT and Harvard
Primer: Agentic AI for Scientific Discovery (Part 1)
George Karniadakis, PhD
Professor of Applied Mathematics and Engineering, Brown University
Seminar: Agentic AI for Scientific Discovery (Part 2)
Nazanin Ahmadi
PhD candidate in Biomedical Engineering, Brown University
Abstract:
Scientific creativity rarely emerges from isolated computation; it often arises through interactions among ideas, models, data, and researchers. While current agentic AI systems can automate increasingly complex scientific workflows, a more fundamental challenge remains: Can AI systems generate genuinely novel hypotheses and contribute to creative scientific discovery? In this talk, we explore agentic AI through the lens of Scientific Machine Learning (SciML), where data-driven learning is coupled with mathematical structure and physical knowledge. We discuss how physics-informed neural networks, neural operators, and related SciML frameworks can serve as computational engines for scientific reasoning and model discovery. We then present an agentic framework designed to support iterative scientific discovery through memory, information sharing, multi-agent reasoning, hypothesis generation, mathematical and computational validation, and self-improvement. Rather than relying on a single model or agent, the system enables multiple specialized agents to propose, critique, test, and refine candidate scientific explanations. Through examples ranging from gene-expression trajectories to cancer-treatment dynamics, we demonstrate how these systems can integrate data, prior knowledge, and mathematical structure to uncover interpretable models and generate new scientific hypotheses. Ultimately, we ask how Agentic AI and Scientific Machine Learning can move beyond automating existing scientific tasks toward identifying unexplained patterns, formulating testable hypotheses, and proposing new directions for scientific research.
Biographies:
George Karniadakis is an elected member of the National Academy of Engineering, National Academy of Arts and Sciences, and a Vannevar Bush Faculty Fellow. He received his S.M. and Ph.D. from Massachusetts Institute of Technology (1984/87). He was appointed Lecturer in the Department of Mechanical Engineering at MIT and subsequently he joined the Center for Turbulence Research at Stanford / Nasa Ames. He joined Princeton University as Assistant Professor in the Department of Mechanical and Aerospace Engineering and as Associate Faculty in the Program of Applied and Computational Mathematics. He was a Visiting Professor at Caltech in 1993 in the Aeronautics Department and joined Brown University as Associate Professor of Applied Mathematics in the Center for Fluid Mechanics in 1994. After becoming a full professor in 1996, he continued to be a Visiting Professor and Senior Lecturer of Ocean/Mechanical Engineering at MIT. He is an AAAS Fellow (2018-), Fellow of the Society for Industrial and Applied Mathematics (SIAM, 2010-), Fellow of the American Physical Society (APS, 2004-), Fellow of the American Society of Mechanical Engineers (ASME, 2003-) and Associate Fellow of the American Institute of Aeronautics and Astronautics (AIAA, 2006-). He received the William Benter Prize (2026), SES G.I. Taylor medal (2014), the SIAM/ACM Prize on Computational Science & Engineering (2021), the Alexander von Humboldt award in 2017, the SIAM Ralf E Kleinman award (2015), the J. Tinsley Oden Medal (2013), and the CFD award (2007) by the US Association in Computational Mechanics. His h-index is 171 (highest in Applied Mathematics) and he has been cited over 184,000 times. In 2025 he was named by Research.com the number one math scientist in the world. (Speaking from 9 - 9:45 a.m.)
Nazanin Ahmadi is a PhD candidate in Biomedical Engineering at Brown University. Her current work focuses on agentic AI in scientific machine learning settings, with applications in systems biology, systems pharmacology, and pharmacometric modeling.
About MIA:
The Models, Inference & Algorithms (MIA) Initiative at the Broad Institute supports learning and collaboration across the interface of biology and medicine with mathematics, statistics, machine learning, and computer science. Our weekly meetings are open and pedagogical, emphasizing lucid exposition of computational ideas over rapid-fire communication of results.
MIA is hosted by the Eric and Wendy Schmidt Center at the Broad Institute.
Relevant Links:
MIA Website: broadinstitute.org/mia
MIA YouTube Playlist: broad.io/MIAPlaylist
Brown University CRUNCH Group: https://www.brown.edu/research/projects/crunch/
Copyright Broad Institute, 2026. All rights reserved.


![Who Did the Analysis? Editors on Disclosure and Peer Review in the Age of AI Agents
The editors-in-chief of Science, Nature, Cell, and NEJM AI on what journals now require when AI does the analysis, and who gets the credit.
Who Did the Analysis? Editors on Disclosure and Peer Review in the Age of AI Agents
A BroadHacks Event | September 22, 2026 | Broad Institute
AI agents are now part of how researchers analyze data, write code, and survey the literature. How do priority and credit work when an AI system produces the result first? Standards for AI-assisted publishing are still taking shape and vary between journals.
PANELISTS
Holden Thorp, Editor-in-Chief, Science
Magdalena Skipper, Editor-in-Chief, Nature
John Pham, Editor-in-Chief, Cell
Arjun (Raj) Manrai, Senior Deputy Editor, NEJM AI; Harvard Medical School
Moderator: Stan Bratchikov, Broad Institute
Topics: disclosure, reproducibility, code review, peer review at scale, licensing, and credit in the age of AI.
*CHAPTERS*
0:00 Roadmap of the discussion
2:01 Panelist introductions
6:42 *Q1: What should PIs write in methods?*
7:33 Pham (Cell)
8:08 Skipper (Nature)
10:50 Thorp (Science)
12:51 Manrai (NEJM AI) [1]
15:02 Skipper: how the community adapts to new reporting tools
16:10 *Q2: What is disclosure for?* Thorp on scientific gotchas
18:17 Skipper: how we can fail as a community [2]
21:50 Pham: do scientists really need journals?
24:17 Skipper: how publishing has changed with AI
25:41 *Q3: The code blind spot* [3]
26:28 Pham: what stops editors and reviewers from checking code
28:01 Thorp: daily pitches and current rejection rates [4]
30:03 Skipper: whats most valuable to check at submission
34:05 Manrai: a system that makes review 17 times faster [5][6]
41:40 *Q4: The Millennium Prize controversy*
42:15 Skipper: Do not upload peer-reviewed research to chatbots
43:50 Skipper: credit in the controversy and its ramifications
46:19 Thorp: how to not get scooped by the internet [7]
49:16 Pham: AI doom and gloom, and why humans do research
50:58 *Q5: Peer review monoculture in a world of two AI models*
51:21 Skipper: humans secret advantage over AI
55:20 Thorp: the Olive Garden problem and AI research
57:40 *Audience Q1: How will publishers handle the volume of AI analyses?*
59:08 Thorp: the analysis storage problem
1:00:00 Manrai: what journals are still hungry for [8]
1:02:27 Pham: the inevitability of trust in research
1:03:43 *Audience Q2: AI editors, and when to stop reviewing*
1:04:08 Thorp: an exciting and terrifying AI future
1:04:42 Skipper: lessons from 30 years of peer review
1:06:31 *Q6: Integrity and self-correction*
1:07:08 Skipper: correction as a driving force of science
1:08:59 Pham: can submissions still be checked at this volume?
1:09:55 Thorp: Sciences publication rate in the age of AI
1:11:16 Pham: Elseviers 153 new integrity hires
1:12:07 *Audience Q3: A special license for the age of AI?*
1:12:25 Skipper: the future of licensing in publishing
1:15:32 Thorp: the issue with CC BY licenses
1:17:12 Skipper: the nuances of small print
1:19:50 *Q7: Who gets the credit in the age of AI?*
1:20:21 Pham (Cell)
1:21:21 Skipper: credit comes with responsibility
1:22:45 Thorp: when LLMs commit crimes, there are no consequences
1:24:18 Manrai: why responsibility is a good thing
1:25:20 Closing remarks
KEY QUOTES
17:40 We have enough gotchas in our life. – Thorp
33:27 No system and no individual is infallible. – Skipper
47:57 The next day I send kids home with a bag of weed and tell them not to smoke it. – Thorp
50:03 We need to make sure that what we are doing does not train us out of jobs as scientists. – Pham
54:30 So what that we have an answer? We dont really understand the journey that led us to it. – Skipper
55:12 Research, just like art, is a human endeavor. – Skipper
1:21:20 It is absolutely essential to think about credit as one side of the coin, where accountability is the other. – Skipper
REFERENCES
[1] 13:57 Nature News & Views (Manrai): https://www.nature.com/articles/d41586-026-02125-z
[2] 21:17 Nature manuscript checklist: https://www.nature.com/documents/nature_3d_ms_checklist.pdf
[3] 26:00 James Zou paper: https://www.nature.com/articles/s41586-026-11044-y
[4] 29:30 Proofig: https://www.proofig.com/ | iThenticate: https://www.ithenticate.com/
[5] 36:12 NEJM AI Fast Track (human+AI review): https://ai.nejm.org/doi/full/10.1056/AIe2501175
[6] 40:04 Liang, Zou et al., LLM feedback on research papers: https://ai.nejm.org/doi/full/10.1056/AIoa2400196
[7] 47:40 Nature on AI text in research papers: https://www.nature.com/articles/d41586-026-02551-z
[8] 1:02:10 NEJM AI Datasets, Benchmarks & Protocols: https://ai.nejm.org/browse/ai-article-type/datasets-benchmarks-protocols
WATCH NEXT
Our previous talk with James Zou: https://youtu.be/SjJtoZFO7Ng
Holden Thorp with James Evans: https://youtu.be/vjmsrDBQyYc
More: https://www.broadinstitute.org
© Broad Institute, 2026. All rights reserved.
#PeerReview #AI #AIinScience #ScientificPublishing #LLMs #Claude #ChatGPT Who Did the Analysis? Editors on Disclosure and Peer Review in the Age of AI Agents](https://i.ytimg.com/vi/Q04PAohvSYU/mqdefault.jpg)







