Specifically, the idea of source credibility just doesn't apply straightforwardly to LLM interactions. When I get info from another person I can ask: how does the content relate to their expertise? Their lived experience? Their interests and biases? 8/

Without considering the sources of claims, efforts at epistemic vigilance become play-acting. With LLMs, we don't know the sources. Excellent thread from @mjcrockett.bsky.social

Every time you get COVID-19, your risk increases for Long COVID, organ damage, disability, & more, & there’s no way to know which infection will be the one to cause it! Protect your lifelong health by protecting yourself from COVID-19. Learn more about Long COVID: www.cdc.gov/long-covid/a...

“Almost 5 years later, #LongCovid has had one of the fastest rises in diagnoses and become one of the most economically disruptive chronic conditions in modern medicine… Another analysis… put the annual economic toll of LC at $1 trillion” — @jasongale.bsky.social www.bloomberg.com/news/feature...

The US NSF has put out a "dear colleagues" letter that is essentially a yard sale list for @ncar-ucar.bsky.social It asks questions such as, who'd like to take over some of NCAR's programs? Is anyone interested in commercial use of the Mesa Lab? If you have thoughts, send them to [email protected]

What is your most galaxybrained thought about why (co)homology turns out to be so important in almost everything?

In this amazing multidisciplinary collaboration, we report our early experience with the @openclaw-x.bsky.social ->

It was a preset list of questions, so no time for audience questions, unfortunately. I am running a session later today based on this: www.peter-keep.com/2026-02-06-g...

It never ceases to confound me just how many folks think the point of essays is to provide teachers with essays. That teachers just like, collect them, or need them for sustenance or something. Rather than the point of them being the *actual fucking act of learning how to think about shit*

AI is going to trash every single way that we had modernised university teaching and send us back to handwritten exams and vivas for every assessment. Which will bring back every problem with those assessment models (and make us seem even more out-of-touch even as we just try to do meaningful work).

Screen reader users can navigate from heading to heading. If HTML headings are used correctly, this allows users to get a basic outline of content and find the information they’re looking for. The highest heading level is <h1> and <h6> is the lowest. developer.mozilla.org/en-US/docs/W...

Is this bad

This is so important. Anytime savings from AI are not going to accrue to the worker. And because many overestimate time savings, some workers are likely going to end up working more than they did before.

BLEAK. Bleak. I’d been paying attention to Columbia selling out its governance to the Trump admin, not so much its stance on AI. www.columbiaspectator.com/opinion/2026...

"large language models such as ChatGPT were consistently advising women to ask for lower salaries than men in recruitment processes,... AI tools already in use by more than half of England’s councils were downplaying women’s medical conditions, potentially resulting in unequal care"

I play this game with my kids, and I have questions about the math. Each player gets a grid to fill with Tetris blocks. Before you start, roll 7 dice and block off the grid spaces you roll. You use your pieces to fill the rest. How did they set up the dice to guarantee you can fit everything in? 🧮

When I’m two weeks behind on my grading.

Incredible to me that all but two major tipping points have uncertainty intervals that dip into heating under 2°C—w coral reefs, the basis for human systems that affect billions, already tipped under 1.5C of heating—and people are out here saying temperature goals for halting heating don't matter.

Hahahaahhaa hell yeah. The future rules! THE FUTURE RULES!

Huge pet peeve: No, “Harvard University” didn’t publish a study, and nor is it proper to say that it is “according to Nature.” Universities don’t publish studies, professors do. And unless it’s an editorial, journals don’t say things either—authors do.

Still can't get over how there's a whole civilization with a capital city at least as big as Pompeii buried deep under miles of volcanic rock from the biggest eruption in like 10,000 years on Lombok in Indonesia and all we know about them is that they existed till they were destroyed

This is really easy, I managed it in less than a minute, so anyone else can.

PSA for Overleaf users: Overleaf has a single "Disable AI features" button if you go to your account settings.

People talk about this stuff in terms of how much it'll suck for gaming and I hate to inform you how many MRI machines, X-Rays, all sorts of other vital machinery are just a Windows IoT Edition PC in a beige plastic box with a big magnet/accelerator/whatever attached.

I appeared as an expert witness before the Joint Committee on AI at the Houses of Oireachtas (parliament of Ireland) to discuss "AI: truth and democracy" this morning. You can read my opening statement here: www.oireachtas.ie/en/publicati...

Question from the 7yr old: Are there types of math questions where the answer is another math question? We said yes, and he wanted examples, so I tried to explain a little of what I remember from Gr12 calculus, but I’m sure there are better answers out there on #edusky #mathtalk #mathsky

Also please don't give a general summary of the text in alt text. I've seen this quite a few times lately. Give the full text. Just copy and paste it. Easy to do. You are trying to give the person who needs alt text as close as possible the same experience of the pic as someone who doesn't need it.

NEW: ChatGPT and other chatbots are feeding users' fixations on real people, reinforcing users' delusional ideas and obsessions about others as they descend into unwanted harassment, stalking, and/or domestic abuse — traumatizing victims and profoundly altering lives. futurism.com/artificial-i...

PSA to my fellow Zoom users -- it's automatically turned on a bunch of A.I. features without asking you✌️ to turn all 32 of them off, go to "Settings" --> "My Account" --> "View advanced features" it'll open up your browser and list a SHITTON of A.I. features, many of which will be on (scroll!)

A three-decade-old open problem of mine just got solved! Matt Kovacs-Deak, Daochen Wang and Rain Zimin Yang showed that if a Boolean function f is exactly computed by a low-degree rational function, the quotient of two polynomials, then f has low decision-tree complexity. 1/2

Preparing for my first lecture (Algorithm Design), I added a slide this year. Can't really not address it.

Hybrid CS Theory Seminar this Fri 2026-02-20! We're excited to have Udi Wieder (Apple) presenting "Uncertainty Quantification Beyond Calibration" udiwieder.wordpress.com www.colorado.edu/cs-theory/th...

“.. ‘many system vendors will go bankrupt or exit product lines due to a lack of memory. Mobile phone production will be reduced by 200-250 million units, and PC and TV production will be significantly reduced.’ Yikes.” @pcgamer.com #DRAM www.pcgamer.com/hardware/mem...

SCOOP: Alpha School, an “AI-powered private school” that heavily relies on AI to teach students and can cost up to $65,000 a year and has been praised by the press and the Trump administration, is AI-generating faulty lesson plans that internal documentation find sometimes do “more harm than good."

oddly straightforward ai ceo

- Major health events — defined as medical claims exceeding $100,000 annually — are now about twice as common as they were five years ago. www.uhc.com/news-article... 1/2

This is why I have a rule that if someone mentions books on social media, I have to go read for st least 15 minutes.

IBM announced they are tripling hiring of entry level workers after their AI experiment failed lol. "We are tripling our entry level hiring, and yes, that is for software developers and all these jobs were being told AI can do." - Nickle LaMoreaux, IBM Chief HR Officer See. Told you.

A complication of AI-assisted coding: developers (even if they’re being careful, checking everything, etc.) are moving too fast to develop mental maps of their projects and remember their design and architectural choices, making it harder to explain those choices or make changes in the future

The "AI revolution" in science is going so smoothly that my instititutional library is having to deploy multiple technical protection measures against "poorly behaved Gen AI crawling" just to maintain its timely service library.ed.ac.uk/research-sup...

Amazing. In all seriousness, no notes

My employer, Univ. Colorado will pay OpenAI $2M/year under the banner of “equity”. That’s 54 full scholarships / year. Plus, our IT will be able to read our chatGPT logs, and our chats can be requested under public records law. No thanks. Surveillance is not equity www.axios.com/local/boulde...

no, it simply isnt. People are talking about capital using gAI as an excuse to conduct mass lay-offs, about students losing learning because theyre being lied to about gAI in education, about the mass theft of the work of artists and authors to train LLMs, about CSAM and nonconsensual sexual images.

"...why AI is so dangerous: people without expertise in a domain feel they know more than they do." 🎯

anyway, what really matters here is that a lot of big companies will stop doing user testing with real humans and instead depend on these AI simulations. The machine will dictate the market and, should we conform to the market, we will ourselves become more like the machine. Unsubscribe.

Last night: wonderful talk by @timnitgebru.bsky.social to a packed house in the Chancellor's Auditorium. This morning: CU system announces $2M chatbot deal with OpenAI, full of misinformation, without consulting shared governance despite impacts on our "principal responsibility". www.cu.edu/gen-ai

I’m so proud of the students in my AI & Society class! They honed right in on this sentence from the university’s FAQ about their new ChatGPT site license, and had a LOT of questions. So do I! I promised I’ll try to find out what I can.

The list of accepted papers at #STOC2026 is out: acm-stoc.org/stoc2026/acc... Congratulations to all authors!

Sighted users often use bold or large fonts to create the appearance of headings in documents. People using screen readers have no way of understanding these visual cues. Use heading styles from the styles menu to correctly format headings.