Group meeting

Monday 29 Dec 2025, 14:00 → 15:50 US/Pacific

405 (Weniger)

14:00 → 14:20 Ethan's report

Speaker: Ethan Muldoon (Oregon State University (US))

Pushed update to do_merge that checks output file contents and makes sure they at least have SOMETHING in them.  It should work for root, hdf5, and tar files, otherwise it just skips the check.  

Still working on my big update that caches metacat and rucio query results so the merge submit script can resume from where it left off in case of problems with the databases.  

Will talk to Kirby about getting a collaboration meeting time slot to talk about merging.  

Need to add some documentation explaining the config file organization.

14:20 → 14:40 Sean's report

Speaker: Sean Gilligan (Oregon State University)

Models by playlist - compare individual vs. trained over whole thing.   

 

14:40 → 15:00 Noah's report

Speaker: Noah Vaughan (Oregon State University)

15:00 → 15:20 Heidi's report

Was able to build phlex

Speaker: Heidi Marie Schellman (Oregon State University (US))

Computing newsletter for December

Phlex install

15:20 → 15:40 Paper of the week - OpenAI and sciene

Questions to consider in reading this paper

1) Several of the authors work for private companies and probably own, or will own, significant # of shares in those corporations. Should they have disclosed their financial conflict of interest (in addition to their affiliations).

2) Are the results found as “novel” as advertised. I read the first example (section 1.1) and came to the conclusion that it seemed a trivial extension of an existing theorem with a proof that one might expect at the senior or beginning grad student level.

2 b) And, how ethical is it to state without including the resulting derivation and proof

"We end this section by noting that our internal models, which can think for a few hours,
were able to derive the optimal bound 1.75/L from scratch (i.e., without providing v1 of the
paper in context, but simply asking the main question studied here directly).”

3) It would be useful for everyone to study section 2 which goes over ways in which chatGPT can be used (and misused) for literature searches.

Speaker: Heidi Marie Schellman (Oregon State University (US))

Early science acceleration experiments with GPT-5