Gary Lupyan

jfc.. I get that LLMs are being used by millions with no background in cogsci/cog neuro/machine learning.. But if they are using LLMs to have meaningful conversations and they think we truly understand how they work (i.e., mech interp is solved), what must they think this understanding looks like?!?

It’s interesting to consider exceptions. I think people are generally open to the idea of other animals feeling pain in himan-like ways

it’s so frustrating to watch smart people not get this.

putting ”inteligence” on an ordinal scale in this way is problematic. chatGPT does a great job writing code for complex ggplot graphs from data and my descriptions of what I need. the usefulness hinges on my ability to check the code for the occasional errors but does that make me more intelligent?

As if irony isn’t already dead, the Hatch Act violation on the DOJ website, which blames Democrats for the shutdown, appears on the page advising DOJ employees they are subject to the Hatch Act. joycevance.substack.com/p/a-note-abo...

They secretly want to mate with us

...it went into crisis mode... This is just an excerpt of a much longer bout of hand-wringing. It then proceeded to forfeit the game and apologized for a "flawed experience".

Problem: The hashes it then spit out do NOT match SHA256! After a few more exchanges where it tried to self debug and offered reasonable possibilities (invisible characters), it kept assuring me that its hashes were correct. I then accused it of cheating. And then...

I protested... and it offered a clever solution!

It's not biting... Its solution is that it will make a choice and then reveal it to me.

A wild exchange with Gemini Pro 2.5. In preparing some examples for my seminar (cogsci-llms.netlify.app) I remembered @randomwalker.bsky.social old tweet about playing Rock paper scissors with LLMs. Decided to see what it's like now ....

Easy fix. Just replace them with people more incompetent than Hegseth. That way they'll have proper respect. 🫡. 😥

Damn! Gonna show this to our 8 year old so he can have something to aspire to ;)

Thanks for engaging! We were working w/ a strict word limit, so couldn't go into lengthy pro/cons. I do think the hard-line nativist position re lang learning (which has been dying for decades within cogsci) *is* dead; @spiantado.bsky.social is right (though not about the lang & thought bits ;)

I love that post! (passed it on to Blaise Aguera y Arcas who said "that is a bomb tweet :)"

if you join @jorge-morales.bsky.social's lab in this position, you also get to work with me and my group! 2 communities for the price of one, what a deal!

seriously though, Jorge is an outstanding mentor, scientist, and thinker, and you'll learn a lot from working with him!

it really is an amazing time (if one puts aside all the other awful stuff). Patterns all the way down...

Serious scholars have examined what happens when we change the number of H1-B visas issued.

Cities that get more H1-B immigrants subsequently see the wages of natives *rise* substantially.

Skilled immigrants bring new ideas, fill labor shortages and make us all more productive.

How important is language for human-like intelligence? We use language to communicate our thoughts. But is language merely the expression of thoughts, which are themselves produced by other, nonlinguistic parts of our minds? Or does language play a more t...

A writeup to go w/ our CogSci 2025 symposium arxiv.org/abs/2509.15560 (in print hopefully soon). We argue that it is not a coincidence that neural net breakthroughs came from training on natural language, and there are important lessons here for understanding human intelligence.

The Soviet system of propaganda and censorship existed not so much for the purpose of spreading a particular message as for the purpose of making learning impossible, replacing facts with mush, and handing the faceless state a monopoly on defining an ever-shifting reality.

My quote of the day

Masha Gessen

Oh, totally! The general point is 💯. Just wondering about the specific psychology behind these overestimates (30% for p(ppl live in NYC)!?. As you wrote, it's not the case for *all* questions, so what's happening here? E.g., is there a subset who give 50% estimates b/c you either do or you don't?!

Cognitive scientist here. None of this is surprising to us. But what’s the baseline? Human reasoning/decision making is not probably correct after all.

Hmm, not seeing it. I wonder how many are actually guessing at random. What they would say if asked to justify their response?

Great writeup! Theres a lot that hinges on the assumption that someone who doesn’t know the answer just guesses randomly. But isn’t the more reasonable assumption that they anchor their response in a similar question to which they know the answer?

This is Hollywood's McCarthy Moment.

This is the litmus test.

Learn from history.

Who will step up and fight?

Who will bend the knee and be complicit?

History will remember.

Psychologists, apply! Apparently we've mainly gotten CS applicants so far. I'd (selfishly) love to have a psychologist. Interests in AI in relation to e.g., implicit bias; cultural influences; fairness; perception; learning & memory; decision making etc. would be a good fit!

Assistant Professor of Psychology - Madison, Wisconsin, United States Current Employees: If you are currently employed at any of the Universities of Wisconsin, log in to Workday to apply through the internal application process.Job Category:FacultyEmployment Type:Regula...

Come be my colleague! UW-Madison Psychology is hiring! Computational Neuroscience and/or Cognitive Science, with an AI emphasis. jobs.wisc.edu/jobs/assista...

RFK Jr does not just reject vaccines. He rejects science and must step down | Bernie Sanders Kennedy is unfit to be our nation’s leading public health official because he rejects the fundamental principles of modern science

⬇️

www.theguardian.com/commentisfre...

Yes, quite.

Image of labubu doll labeled labubu next to image of spiky labubu doll labeled lakiki

::slowly stands while clapping::