[ESIP-all] AGU Session: Test and Validation of AI Models in Earth and Space Science

Hamed Alemohammad HAlemohammad at clarku.edu
Mon Jul 17 10:41:24 EDT 2023


Dear All,

We invite you to submit your abstracts to the AGU23 Fall Meeting session on Test and Validation of AI Models in Earth and Space Science. We think this may be one of the more timely and important topics in scientific information technology today.  Much of the hyperbole about AI models is driven by examples of irresponsible or improper use of these tools and science community needs to have a broad-based discussion of how to counter this hype by reasoned validation of these models so as to make them trustworthy. This comes more sharply into focus when they are re-used by others outside the original development team. We think you have something important to contribute to the conversation. Please consider submitting an abstract to this important session.

The AGU abstract submission site:
https://agu.confex.com/agu/fm23/prelim.cgi/Session/190454

Abstract Submission Deadline: Wednesday, 2 August 2023

Session description:
AI Models are more than just ‘ChatGPT’.  New classes of models are emerging to enable re-usability.  Pre-trained general AI models, including Foundation Models, can provide a base from which new models can be generated to address specific investigations quickly and efficiently.  Re-usability requires, among other things, the publication of AI models, including the code, the training dataset, test datasets, and the results of validation campaigns. This is particularly true for Foundation Models, intended for re-use in many different ways.  The science community must insist on thorough validation before accepting new models for re-use. This session is to explore experiences model developers and science researchers have had with model validation. Topics would include approaches to create benchmarks for AI models, methods to validate AI models and to report results, and what elements the science community needs to create confidence in a model.


Sincerely
Hamed Alemohammad, Clark University
Mark Carroll, NASA Goddard Spaceflight Center
Rahul Ramachandran, NASA Marshall Space Flight Center / NASA IMPACT
Mike Little, NASA Goddard Spaceflight Center

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.esipfed.org/pipermail/esip-all/attachments/20230717/b86dbc6b/attachment-0001.htm>


More information about the ESIP-all mailing list