Should also be done by moderate centrists, economic nationalists and policy hawks, labour-oriented leftists, defense analysts, and governance experts; all for very different reasons. Just not being done enough. China’s AI trajectory has started to take a quantifiably different form from the US [+]

“Denoising with retrieval” is a rather cool idea. Instead of using search as part of an internal monologue that always moves forward, use it diffusion-wise to recognize noisy areas of your draft that need more input. If this is SOTA deep research, that’s an impressive application of diffusion.

Worse news: there are uncountably many undecidable problems. Computers have been a scam this whole time

nobody will remember: - your salary - how “busy you were” - how many hours you worked people will remember: - the twenty-first night of september - changin' the minds of pretenders - chasin' the clouds away

Aaand ... we have a full-circle connection back to psychology!

It would be fun to map the network of inter-field connections created by quote posts here. For instance, here the chain starts with a psychologist describing how hard it is to operationalize "sleep"! It passes through comp. neurosci and is now being QTd by an info sci prof obsessed w/ benchmarks.+

Great point. Benchmarking has become the name of the game in machine learning. You can effectively add to or subtract from the definition of intelligence by introducing a new benchmark or critiquing an existing one, and convince a significant portion of the community that you have a valid point.

Some TPOT and e/acc folks have started popping in again lately. Hi! Your unhinged posts and chatbot romances are essential to our healthy AI ecosystem, so please check out my old post on constructing your feed to make bsky enjoyable.

Watching some of these evangelical influencers giddily describe how they’re prepping for the big day certainly gives me flashbacks to being a kid terrified his parents had been raptured & he left behind anytime they were late home from work in a pre cell phone era+

Aujourd’hui, le Canada reconnaît l’État de Palestine. Ma déclaration : www.pm.gc.ca/fr/nouvelles...

Two posts here that could have a fruitful conversation are, on the one hand, @ai4geo.bsky.social 's thread on ACI: defining goals for AI that are explicitly not human-shaped. +

“First we came for Jimmy Kimmel, and everyone spoke out, because on reflection he was that normie dude from the Man Show.” “Why did we do that? We gained bupkis. Why didn’t we start with the communists like in the fucking poem?”

AGI and ASI are understandable but actually pretty weird goals for AI research. Both are defined in terms of human capabilities and those capabilities are framed in what's valuable in a human centric society...

There is no world in which it is normal for the president to publicly call upon his attorney general to hurry up and prosecute his political foes. It’s like the Watergate tapes but posted on social media. Let’s get a grip on what’s happening here.

I think a large part of the phenomenology of doomscrolling is that you are absorbing information without communicating in a physically engaging way (ie speaking). So, it’s all “input” and no “output”. This leads to feelings of isolation or helplessness

When souls pass from Purgatory into Heaven they meet neither mockery nor reproach for their detour, but are welcomed into fellowship with open-hearted gladness, and that’s how I’m going to welcome new people to Bluesky, yea, even if I have to bite my tongue so hard that it draws blood.

I think the AI job automation/augmentation dichotomy doesn’t help much. Automation can be a form of augmentation (have the AI write an analyst report so analysts can focus on higher value work) & augmentation can be automation (you may need less workers if productivity goes up)

Look at academia giving me a little hope

Eyvind Earle

How to quantify the impact of AI on long-run cultural evolution? Published today, I give it a go! 400+ years of strategic dynamics in the game of Go (Baduk/Weiqi), from feudalism to AlphaGo!

`A caricature 1940s pinup girl in a beach setting` #midjourney

I get why everyone is dunking on this, but, having trained a very simple model on the ~3 million words I’ve written over the last 25 years, it’s actually very interesting to explore what’s there, to see what’s surprising or where I’ve changed my mind, or where my opinion diverges from its output.

The media is doing a bad job of covering this $100k H1-B “fee” (actually tax, since it is not implemented by an agency to cover processing costs). Taxing via EO looks extremely vulnerable to legal challenge, and should be treated as stochastic disruption not durable change.

I know it’s boring to announce that you canceled Disney+. But they make cancellation such a long, labyrinthine process that it’s natural to boast about it

The media is doing a bad job of covering this $100k H1-B “fee” (actually tax, since it is not implemented by an agency to cover processing costs). Taxing via EO looks extremely vulnerable to legal challenge, and should be treated as stochastic disruption not durable change.

The people on this site crave papers. You can post a random excellent paper from 1978 on any topic and it'll blow up a little bit. You can fill the gap!

I'm not trying to tell you it's not going to suck but this is more or less what the path to most of the good endings looks like: 1) isolated, unpopular president with few allies 2) entirely transactional and lightly committed collaborators 3) deeply committed domestic and international opposition

What we are witnessing is an outright abuse of power. This administration is attacking critics and using fear as a weapon to silence anyone who would speak out. Media corporations — from television networks to newspapers — are capitulating to these threats.

Trying to keep my professional chill but I’m SO excited Carnegie Mellon is launching a cluster hire in computational humanities—MULTIPLE JOBS! 1. Asst Teaching Track Prof in Computational Humanities - apply.interfolio.com/173622 2. Asst Tenure Track Prof in CH - apply.interfolio.com/173626

"chuds are just mad that their shit doesn't sell anymore" this looks very naive now in light of a near-total alliance of media outlets and platform owners with reactionary politics

My genuine memory of myself is that I was roughly this, which makes me think if I went back in time I'd actually find myself actually calling for the guillotine for anyone who failed to capitalise the "B" in "Black people".

I was pretty arrogant about it. In hindsight, the online left kept making excuses for corporate power by minimizing their power to censor as less real because it is not government censorship. You can't square that with a critique of say, corporate media.

"Deplatforming!" was a very oddly chosen rallying cry for a left that never, in material fact, owned most of the platforms.

If, hypothetically, your stance in 2015 was "I'm not super worried about attacks on speech from the left, but for the record 'free speech' is still an ideal we are going to need and sneering at it won't age well" — it would appear you had a point.

I needed an image to abstractly imply "pluralism," and as ten-minute brainstorming sessions go, the one that produced this was not bad. (It's in alt-text.)

DeepSeek published about R1 in Nature www.nature.com/articles/s41... but the supplementary information is far more detailed. it includes, lots of details on everything from datasets, data processing, hyperparameters, and more static-content.springer.com/esm/art%3A10...

Just got to Purdue for the Modern Fiction Studies ~Cultural AI~ conference! Excited to present a new paper, coauthored w/ @aaronrhanlon.com, that LLM outputs constitute a new genre we're calling stochascript & learn what others are working on. Also excited about this super cute PPT template 🤓

When it comes to text prediction, where does one LM outperform another? If you've ever worked on LM evals, you know this question is a lot more complex than it seems. In our new #acl2025 paper, we developed a method to find fine-grained differences between LMs: 🧵1/9

This came up in my class yesterday. We read @oms279.bsky.social’s wonderful paper “From Codebooks to Promptbooks” (highly recommend for teaching), but got stuck on the issue of non-random errors, which is a not a unique issue to LMs. Traditional methods… Paper: journals.sagepub.com/doi/10.1177/...

I’ve noticed one thing in life: some people don’t like change, and persuading them to change is just painful. For example, AOL still has thousands of dial-up Internet customers. AOL is on sale for $1.5B. www.wsj.com/business/dea...

The upside of this is we could be equally shameless. "We're going to convince young men that oligarchs are cringe and talk down to them but Mamdani meets them as an equal. NYT, publish our propagandatorial or you're biased."

We're the only ones into free speech so unfortunately Bluesky is the intellectual dark web now

In our forthcoming paper, John Hummel and I ask what it would mean for a neural computing architecture such as a brain to implement a symbol system, and the related question of what makes it difficult for them to do so, with an eye toward the differences between humans, animals, and ANNs.

You know, after all the saber-rattling and visa harassment, international enrollment in UIUC is *up* from last year. We’ve lost a lot of grant money, but I don’t feel the character and ideals of the school have been compromised. Fingers crossed!

Cool new codex behavior -- it uses a preview software to take screenshots of the front end it's writing, to debug visually instead of just code. Then deletes the screenshots and makes changes or wrapping up.

I'm sticking with the "2027" date for the Great AI Boyfriend Panic of 2027. These are just foreshocks. arxiv.org/abs/2509.11391

🚀 Introducing Fluid Benchmarking—an adaptive way to evaluate LLMs. Inspired by psychometrics, it tailors which questions to ask based on each model’s capability, making evals more efficient & reliable. 🧵

Are you working on a #DH related piece? Are there databases or tools that you think are great for research, but not well-known? The #DO accepts guest pieces on a rolling basis. Submit your work, and give your work visibility #submission #callforpapers #openaccess Link in bio

Gorgeous lift door in 33 N Lasalle, Chicago

Just sharing a laptop sticker that cheers me up when I look at it; got this from @thegurglingcod.bsky.social