CJ
Assistant professor with too many opinions. Texpat, politics academia, gay stuff, anti-carbrain. Not an AI brain genius guy. "Train Linguist." 🇵🇸🏳️🌈🇺🇦🏳️⚧️
"See you divas on the streets."
E me aperta pra eu quase sufocar
BayesForDays@lingo.lol on Mastodon.
- Nicest compliment I've gotten in a long time: "it sounds like [...] you are pretty technical"
- Feeling kinda bitchy that my citation counts are wrong on Google Scholar
- Reposted by CJ“We are going to force DHS to follow the law” should have been the compromise position before you decided to allow a 175B budget for ICE to pass. Falling back to that now is just weak. Do some fucking rescissions of its funding, you feckless cowards.
- love a good snekdown
- Reposted by CJ"Abolish ICE" is the law-and-order position
- Reposted by CJin other words, we’re still in the middle of abolition, the civil war, the civil rights era
- Reposted by CJI do not regret to inform you that we are going to win.
- Literally one of my HSP reviewers didn't know that "roost" is a word lmao
- Reposted by CJThis must be weird news to see if you’re one of the literally hundreds or even thousands of university administrators who preemptively censored faculty, scrubbed websites, changed the names of centers, etc.
- Trump administration drops legal appeal over anti-DEI funding threat to schools and colleges. This leaves in place a federal judge's ruling that the anti-DEI effort violated the First Amendment and federal procedural rules. (Via AP) apnews.com/article/dei-... #ncpol #nced

- Oh I guess it's 2026 now so it's actually Worse
- Reposted by CJLast year we went to Memphis to document the emissions from the xAI datacenter turbines. They were bad. Even Trump's EPA seems to agree. "A US regulator ruled on Thursday that Elon Musk’s artificial intelligence company had acted illegally by using dozens of methane gas turbines..."
- Ok I'm logging off again I'll be back when I have more fun science to share
- LLMs are making for academic monoculture and the time is nigh for dropping them entirely imo
- Just got sent yet another paper that is a computationally expensive but trivial extension of literally hundreds of other analyses that have brought us no closer to understanding how the mind works
- Can y'all please stop doing this to me. I don't publish slop and it's not fair to hold a field of good researchers to slop standards
- We wrote a thing -- showing you don't need LLMs to model language production dynamics like the tendency for speakers to reduce predictable words. All you have to do is better model how speech rate varies depending on where a word is and how long the utterance is. arxiv.org/abs/2512.23659
- We explored this problem by re-analyzing two recent papers, showing that (1) small-scale language model surprisal outperforms long-range surprisal, and (2) bigram transition probabilities do better than LLMs. Then we conducted two new studies -- showing the same thing as (2).
- Beyond better capturing utterance structure, one of my favorite contributions of the work is that my students pushed me to expand the work to corpora outside Mainstream American English. We analyzed both Buckeye (a classic!) and all of the CORAAL data and show phrase-sized probabilistic reduction
-
View full threadAnd since n-grams are all you need, this provides some nice converging evidence that speakers probably use very local representations to pace their utterances, say by retrieving phrases. We would love comments!
- I love that all the surprisal bros have suddenly discovered mutual information like the collocations people haven't been working with this for the last thirty years
- It genuinely feels like them rummaging through a set of like 4-5 measures that we mostly already ruled out in the 70s and that really don't advance our theory of anything
- I feel like such a jerk about this but it's really quite exhausting. But idk I'm not a rockstar or whatever
- It is always so exciting when people who have publicly anguished about transitioning finally transition
- Not including random effects by items when you're studying effects for which different items are known to show vastly different patterns of results is scientific malpractice in 2025
- It's ok to do this if people think you're brilliant though
- I'm not on here anymore and I think everything is just as horrifying as it was. I'm on mastodon
- what if i look away for just a second and everything is different
- Insane to do a coup in broad daylight jesus
- you're not supposed to be able to find out about these things! isn't it the whole idea to like, lie about it
- these moments like waves
- new benchmark: any profile that looks like a prompt is a bot
- god the botnets are swarming this fucking sucks
- i just got followed by a bot account called purple house photos????????
- "The most important thing about AI isn't its technical capabilities or limitations. The most important thing is the investor story and the ensuing mania that has teed up an economical catastrophe that will harm hundreds of millions or even billions of people." pluralistic.net/2025/09/27/e...
- Reposted by CJThe line people ran with when Harvard was under the crosshairs was that Ivy League schools are just training grounds for the elite, and the actual fight for racial equity would be won in regional, state, and trade schools where “real” Black people could improve their economic outcomes.
- it is very funny to me to see the "genius" level folks in my field just re-hash older papers and present the work as if it were their own
- indistinguishable from the snitches at Texas universities in gender studies classes, btw
- At Cornell University, an eminent prof, whose work I was reading in graduate school got suspended b/c an IDF veteran grad student enrolled in his class for the express purpose of disrupting it. We are all cooked. www.thenation.com/article/soci...
- Would UB protect me?
- i am so tired of hearing the voices of other people lol
- Reposted by CJEveryone in the kill chain should be prosecuted but we have an opposition leadership that thinks this is a distraction from funding rural hospitals in +70-80% trump country
- Reposted by CJTired: Salt Fat Acid Heat Wired: Harissa Explains It All
- Reposted by CJ
- lightness
- Reposted by CJI encourage you all to add your voice to protectiing the U.S. National Fungus Collection! #Mushrooms #FungiFriends
- The federal government lists "indica" (sic) of violence and I'd like to know whether fascists instead have "sativa" of violence
- These guys (or their language models) do a lot of drugs is what this typo tells me
- Bikes are freedom
- There's a lot of history in this photograph ( en.wikipedia.org/wiki/United_... )