Nikhil Krishnaswamy
@nikhilkrishnaswamy.bsky.social
Assoc. Prof. of Computer Science, Colorado State University. I hate technology. https://www.nikhilkrishnaswamy.com/ #AI #NLProc Add me to your war planning chats: NikhilKrishnaswamy.77
(Not so?) hot take: requiring evaluation on the largest (i.e., most expensive) "frontier" models is irresponsible reviewing. It implicitly requires giving money to large corporations in order to do research and explicitly disadvantages entire regions without the financial resources to do so
The public sees the corporatization of universities as a leading problem but the public also routinely gets snookered into thinking that public institutions should be run by businessmen 🤷♂️
Opinion: Administrative bloat is no longer an inside-baseball complaint. Its a public-trust issue for higher ed. Administrative and support payroll grew 88 percent from 2010 to 2024, outpacing revenue and instructional salaries, according an analysis of federal data. https://chroni.cl/46urF3W
Watching your neighbors and all your friends have a killer party over the fence that they only held just to not invite you. This post is about geopolitics.
Aka why the Democratic Party should've been listening to me for the past six funking years
"On the merits, it’s just good policy. Anyone opposed to America turning into a society governed by a ChiCom-style social credit system should be opposed to cameras. On the politics, it’s even better." It's just right there for the taking:
Real frustrating when you work the math out right but you still have to do hyperparameter tuning...
"Corporate Lysenko" is the name of my new band. Why that? No reason.
Hey buddy you got any breakthrough proofs you’re working on? No reason, me and my large corporation are just curious
New plan: keep the big AI labs distracted with whether the proof of Navier-Stokes is real or not while the rest of us do the rest of science
There are a number of ideas going around in AI that are inherently very interesting but unfortunately not many that stand a chance of extricating us from our current scientific and socioeconomic predicament
My 3yo is getting very good at identifying aspen trees. Unfortunately his way of doing so is "the trees have eyes"
Spain beat Portugal so I guess they get control of Asia now #worldcup #treatyoftordesillas
12 years of school, 4 years of college, 6 years of grad school, 3 years of postdoc, 6 years on the tenure track, all for a job title that can still be abbreviated Ass. Pro.
Yard sales are just trick-or-treating for adults
Want to beat China in the "AI era"? Invest in the humanities.
China’s universities cut thousands of ‘obsolete’ arts degrees in AI overhaul
Nearly a third of China’s university programmes have been impacted
independent.co.uk
I've been to 3 #AI conferences in 4 weeks. The one week I wasn't at a conference, it was a project site visit. Turn on @npr.org and they're talking about AI. I'd like to be done now. #LREC2026 #AAMAS2026 #CVPR2026
When the CSU doing something crappy isn't your CSU
"The AI-centrist framing of 'responsible use' does not constitute a legitimate plan to protect the university and its students against the potential harms of this technology. Nor does it address the social and environmental impacts of creating and operating large language models."
Bonus panel here: www.smbc-comics.com/comic/ai-17 Also last day to get my new book(s) on kickstarter! www.kickstarter.com/projects/wei... #smbc #comics #ai
AI is still just people. When you ask "AI" you're not hearing from "AI", you're hearing from the people who control the AI.
New paper in Nature. The more a government controls its domestic media, the more it dominates AI training data, the more pro-regime outputs we get from AI. By scraping the open web, LLMs are unwittingly laundering state-coordinated narratives into seemingly objective answers.
Today I'm going to help my neighbor DIY fix our shared fence. He said he asked "my friend Gemini" how to do it. I didn't give him the spiel about what I do every day. We'll see how this goes. #Al #DIY #LLMs #situatedgrounding
Before kids: We will raise them to be critical thinkers, with no ridiculous superstitions. After kids: If you don't eat your dinner and stop screaming I'm calling the Easter Bunny and telling him not to come!!!!!!!
If I hit my neighbor and he steals my wallet, so I pay him money to get him to give my wallet back, what kind of strategic genius am I?
"I’m going to suggest that overreliance on AI is bad for you — which suggests a corollary: Whatever is happening to normal people as a result of AI overreliance has already happened to the ultra-wealthy." Read more from @lopatto.bsky.social: www.theverge.com/tldr/897566/...
Please take medical advice from people who count 1, 2, 3, 3.1, 95, 98, ME, XP, Vista, 7, 8, 10
The Copilot Health waitlist is available to join starting today.
“I like where you’re going with it for sure, it’s just, uh, I was thinking we’d go with something a little more, I don’t know, modern?” said Vice President JD Vance
JD Vance Struggling To Articulate What It Is He Doesn’t Love About Baby Name ‘Rohit’
WASHINGTON—Scratching his chin as he appeared to search for the right words, Vice President JD Vance reportedly struggled to articulate Wednesday what exactly it was that he didn’t love about the baby...
theonion.com
While there are cases where "AI" (however you define that) can be helpful for things like human flourishing and well-being, the emphasis on scale as necessary for bringing them to fruition is inherently totalizing & makes its integration with the machinery of destruction seem inevitable
Cool new benchmark and dataset to appear at LREC 2026. See you in Palma! #NLP #AI #LLMs
Distributed Partial Information Puzzles: Examining Common Ground Construction Under Epistemic Asymmetry
Establishing common ground, a shared set of beliefs and mutually recognized facts, is fundamental to collaboration, yet remains a challenge for current AI systems, especially in multimodal, multiparty...
arxiv.org