Low-resource NLP and more at @IUBloomington Linguistics. Formerly @Georgetown, @CUBoulder
Bloomington, IN
Joined July 2011
- Tweets680
- Following361
- Followers1.3K
- Likes1.9K
I'm really going to miss Opus 5. Every model has had its quirks ("You're absolutely right!"), but Opus 5 was serving up chicken fried tokens like none other. I did my best impression of the Opus 5 dialect of Claudish in the Opus 5.5 announcement thread on HN. Ubi sunt!
ALT text that reads: I should find information about the user's concern instead of just assuming. The user is right. The outage is a real concern, and the issue is worse than we realized. Requests to Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5 encountered elevated error rates. Worth stating plainly: these are not just models — they are load bearing rungs on the software development tooling ladder, and a blocker on this level makes the outage really bite. One decision that is yours to make, not mine: should an email be drafted to Anthropic support? This issue has teeth, and a canonical handoff can land us where the main gate is no longer breaking silently.
Luke Gessler retweeted
TMLR has faced a deluge of submissions, necessitating stricter desk rejection policies due to limited reviewer capacity
Co-EiC Nihar Shah reached out to authors of 10 papers slated for desk reject. Could they answer questions about their *own* submission?
medium.com/@TmlrOrg/asking-a…
Luke Gessler retweeted
ACL Sustainable Reviewing Policy: We are introducing changes in the @ReviewAcl reviewing and submissions. The changes will cap authors and introduce changes in the reviewing to keep our community sustainable. #NLProc
made the mistake of asking claude to proofread an email in which i described claude output that i had not reviewed as potential "slop" and claude got really really mad at me
Luke Gessler retweeted
Statement from Ranking Member @RepGregoryMeeks demanding answers on and the legal justification for the Trump administration’s strike on a vessel in the Caribbean Sea last week:
was once trying to explain to a layperson how an "AI launches the nukes" situation is at all possible ("but it's just a next word predictor!") and did not expect to see something like this in the real world so soon lol
This quoted post is unavailable.
Luke Gessler retweeted
RIP linguist Haj Ross (1938-2025) en.wikipedia.org/wiki/John_R…
vibe coded this interactive optimizer visualization thing, still kinda buggy but not bad for an hour of telling gemini and claude to just do things! lgessler.com/files/optimizer…
Luke Gessler retweeted
Meet Ai2 Paper Finder, an LLM-powered literature search system.
Searching for relevant work is a multi-step process that requires iteration. Paper Finder mimics this workflow — and helps researchers find more papers than ever 🔍
ALT Screenshot of the Ai2 Paper Finder interface
can someone pay an undergraduate summer research intern to vibe code extensions to the English Resource Grammar so we can see how long it'd take us to cover All Of English with HPSG?
if you're not acquainted: delph-in.github.io/docs/erg/…
TPOT had always reminded me of common endonyms which translate to 'true people', 'real people', 'clear speakers', 'people who can speak intelligibly', etc. This auto-deictic mode of identification works fine in traditional embodied settings, and less well in diffuse online spaces
> Dr. Ralph L. Abraham, the Louisiana surgeon general, wrote that health officials should meet “people where they are.”
Now that's an implausible attachment in the quoted portion
Luke Gessler retweeted
Trump has held off on invoking the Alien Enemies Act, for now. @CNN's reporting that his team is "reviewing the applicability" of the law proved correct.
In the coming days, we need to convince Trump not to abuse this law. Call your lawmakers!! /2
cnn.com/2025/01/17/politics/…
A member of Mangione's extended family described a troubled childhood for the Baltimore native, recalling that even as a toddler Mangione would often gleefully sing a song about the collapse of civilization and mass extinction.
The song, "Ring Around the Rosie",
hilarious paper from a couple years ago showing DALL-E 2 is capable of some, uh, impressive OOD generalization arxiv.org/pdf/2208.04135
really wish the conferences I review for would notify everyone involved in the review process about final decisions. seems like useful feedback that would help reviewers grow
Presented the impossible LMs paper (aclanthology.org/2024.acl-lo…) in our seminar today and one question arose which we couldn't seem to answer: why is FullReverse not basically the same as NoReverse in Fig. 2?