Sanmay Das
@sanmayd.bsky.social
CS prof at VT Innovation Campus, Assoc. Dir., AI for Social Impact, Sanghani Center. Professional: AI and society, EconCS, etc. Chair, ACM SIGAI. Personal: Husband, dad, squash player, cyclist. From Delhi to NoVA, via Boston, upstate NY, STL, others...
After having more than one SPC at AAAI produce AI-generated “metareviews” (that were actually “reviews” in the first place) my only remaining faith is that by continuing to layer the hierarchy of reviews further the law of large numbers will kick in in some fashion…
November 18, 2025 at 2:39 AM
After having more than one SPC at AAAI produce AI-generated “metareviews” (that were actually “reviews” in the first place) my only remaining faith is that by continuing to layer the hierarchy of reviews further the law of large numbers will kick in in some fashion…
I think many people are misreading this as being about all cs arxiv papers, not just review/position papers!
November 9, 2025 at 3:27 AM
I think many people are misreading this as being about all cs arxiv papers, not just review/position papers!
Yup, this. And on the second question, I think it’s when ICML and NeurIPS introduced position paper tracks.
November 9, 2025 at 3:08 AM
Yup, this. And on the second question, I think it’s when ICML and NeurIPS introduced position paper tracks.
If you’ve had issues of this kind with cs.CY, you should always feel free to appeal. Mods for the categories are, from my experience, both well-trained and read widely in their areas, and understand the norms of relevant disciplines pretty well!
October 31, 2025 at 7:51 PM
If you’ve had issues of this kind with cs.CY, you should always feel free to appeal. Mods for the categories are, from my experience, both well-trained and read widely in their areas, and understand the norms of relevant disciplines pretty well!
This is also a big part of why there’s suddenly a preponderance of “multiagent systems” — they need modularity for individual tasks within a larger project because otherwise they get muddled by the context window.
October 30, 2025 at 11:37 AM
This is also a big part of why there’s suddenly a preponderance of “multiagent systems” — they need modularity for individual tasks within a larger project because otherwise they get muddled by the context window.
Sadly, it will likely be more important than the famous Conservapedia
October 28, 2025 at 12:15 AM
Sadly, it will likely be more important than the famous Conservapedia
Thanks for an excellent talk! Great work
September 30, 2025 at 9:43 PM
Thanks for an excellent talk! Great work
Lol, will take your word for it!
September 15, 2025 at 11:01 AM
Lol, will take your word for it!
Also they do all this bio and chemistry stuff, but no math! Monotonic! Affine!
September 15, 2025 at 1:23 AM
Also they do all this bio and chemistry stuff, but no math! Monotonic! Affine!
Although, also, mangos are one of the easiest fruits to cut and eat??
July 16, 2025 at 11:15 AM
Although, also, mangos are one of the easiest fruits to cut and eat??
I have to inform you, though, that the flavor and aroma of the Dasheri are unmatched (with citation!) en.m.wikipedia.org/wiki/Dasheri — my local store has flown in boxes of both the last couple of years :)
Dasheri - Wikipedia
en.m.wikipedia.org
July 16, 2025 at 11:14 AM
I have to inform you, though, that the flavor and aroma of the Dasheri are unmatched (with citation!) en.m.wikipedia.org/wiki/Dasheri — my local store has flown in boxes of both the last couple of years :)
Reading this at face value, on top of everything else, “immigration” is a “form of discrimination?!”
July 12, 2025 at 12:26 PM
Reading this at face value, on top of everything else, “immigration” is a “form of discrimination?!”
Ah, I found the post I was thinking of! bsky.app/profile/fili...
Look, I did my undergrad in Brazil, and there were very, very few essays, and take-home assessments in general, and lots of in-class exams. And do you know why? Because cheating was rampant! And instructors knew it!
That is: AI didn’t invent cheating, so we know how ppl adapt to rampant cheating.
That is: AI didn’t invent cheating, so we know how ppl adapt to rampant cheating.
I gotta say I find a lot of AI discourse around higher ed very confusing. "if ChatGPT can write your essays is college even worth it?" did people think math teachers were assigning problem sets because *they* couldn't figure out the answers?
July 10, 2025 at 2:56 AM
Ah, I found the post I was thinking of! bsky.app/profile/fili...
I saw someone posting about college and cheating in other countries (I wish I remembered where/who) and basically I agree with the idea that we need to (in higher ed) go back to exam-based assessment with no Internet access, etc. Which I sort of hate, but, yeah…
July 10, 2025 at 12:26 AM
I saw someone posting about college and cheating in other countries (I wish I remembered where/who) and basically I agree with the idea that we need to (in higher ed) go back to exam-based assessment with no Internet access, etc. Which I sort of hate, but, yeah…
But there’s also this: we don’t let second graders use calculators when learning arithmetic. We don’t give sixth graders access to the Internet when answering reading comprehension questions…maybe we can double down on why what we’re teaching (without LLMs) is important? When it is, that is…
July 10, 2025 at 12:20 AM
But there’s also this: we don’t let second graders use calculators when learning arithmetic. We don’t give sixth graders access to the Internet when answering reading comprehension questions…maybe we can double down on why what we’re teaching (without LLMs) is important? When it is, that is…