[CODATA-international] AI has a metadata problem....

Mark A Musen musen at stanford.edu
Tue Apr 25 16:59:12 EDT 2023


Thanks for opening up this discussion.  I agree strongly that AI models will benefit from the kinds of metadata explored by the CODATA community, and that our work on metadata—and particularly data FAIRness—will be enhanced by taking advantage of methods developed by the AI community in its early work on knowledge modeling.  There is tremendous synergy here that needs to be exploited as we try to capture in metadata the essence of the content of digital research objects.

Here is a link<https://www.nature.com/articles/s41597-022-01815-3> to a recent paper that discusses some of these issues.

Mark
_________________________
Mark A. Musen, M.D., Ph.D.
Stanford Medicine Professor of Biomedical Informatics Research
Professor of Medicine and Biomedical Data Science
Head, Stanford Center for Biomedical Informatics Research
Stanford University School of Medicine, Mail Code: 5702
3180 Porter Drive, Room B106
Palo Alto, CA  94304

Phone: +1 (650) 725-3390
e-mail: musen at stanford.edu



On Apr 25, 2023, at 12:10 PM, Matt Jones <jones at nceas.ucsb.edu> wrote:

Thanks, that was a nice post, I enjoyed it. It struck me as related to the FARR project, which is trying to tackle a similar set of issues as those you outline:

FARR: FAIR in ML, AI Readiness, & Reproducibility (https://www.gofair.us/farr)

They published a recent case study on publishing AI models in physics, and are working on other disciplinary issues as well:

Duarte J, Li H, Roy A, Zhu R, Huerta EA, Diaz D, Harris P, Kansal R, Katz DS, Kavoori IH, Kindratenko VV, Mokhtar F, Neubauer MS, Park SE, Quinnan M, Rusack R, Zhao Z (2022) FAIR AI Models in High Energy Physics. https://doi.org/10.48550/arXiv.2212.05081

Matt

Matthew B. Jones
ORCID: 0000-0003-0077-4738<https://orcid.org/0000-0003-0077-4738>
Director of Informatics R&D, National Center for Ecological Analysis and Synthesis<http://www.nceas.ucsb.edu/ecoinfo>
PI, NSF Arctic Data Center<https://arcticdata.io/>
Director, DataONE<https://dataone.org/> program
University of California Santa Barbara


On Tue, Apr 25, 2023 at 7:01 AM Pascal Heus <pascal.heus at postman.com<mailto:pascal.heus at postman.com>> wrote:
Dear colleagues:
I just published this short article around AI transparency and metadata challenges which might be of interest to you. One of the objectives is to foster a bridge between the AI and (meta)data communities.
https://plgah.medium.com/ai-has-a-metadata-problem-78b30ca1936b
Thoughts, feedback, suggestions most appreciated.
Best,
*P
_______________________________________________
CODATA-international mailing list
CODATA-international at lists.codata.org<mailto:CODATA-international at lists.codata.org>
http://lists.codata.org/mailman/listinfo/codata-international_lists.codata.org

The CODATA International list is for announcement of of activities, events and outputs by CODATA and by other organisations and initiatives.  It is also for discussion of all issues related to data.  It is an open subscription list with only lightweight moderation to remove spam.  Messages posted on the list by third parties do not necessarily imply endorsement by CODATA.
_______________________________________________
CODATA-international mailing list
CODATA-international at lists.codata.org
http://lists.codata.org/mailman/listinfo/codata-international_lists.codata.org

The CODATA International list is for announcement of of activities, events and outputs by CODATA and by other organisations and initiatives.  It is also for discussion of all issues related to data.  It is an open subscription list with only lightweight moderation to remove spam.  Messages posted on the list by third parties do not necessarily imply endorsement by CODATA.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.codata.org/pipermail/codata-international_lists.codata.org/attachments/20230425/cb8edb38/attachment.html>


More information about the CODATA-international mailing list