A PhD is much more than cheap research labour. A PhD is an apprenticeship in research. Today’s students are tomorrow’s research leaders. Students learn how to ask the right questions, how to critique findings and, ultimately, how to take responsibility for the science they produce.
⚛️ Introducing CREATE, a benchmark for creative associative reasoning in LLMs.
Making novel, meaningful connections is key for scientific & creative works.
We objectively measure how well LLMs can do this. 🧵👇
Experiments show that current LLM-based coding agents can generate working code in the short term, but struggle to maintain code quality across many iterations and evolving requirements.
#aicoding #agenticengineering
📄 arxiv.org/abs/2603.03823
AI vision systems mistake a hairless cat for an elephant because they rely on texture and pixel patterns, while humans recognize objects by shape and context.
This gap poses real risks in self-driving cars and medical imaging.
"AI" is not a stochastic parrot.🦜
I wrote this piece a couple weeks ago, but it was hard for me to finish up given AI's role in society and war over the past few weeks. I should share it at some point though. Not perfect, but here it is.
medium.com/@margarmitch...
Fantastic essay: "Why the ATM didn’t kill bank teller jobs, but the iPhone did." davidoks.blog/p/why-the-at...
Key point: "it is paradigm replacement, not task automation, that actually displaces workers." In this sense, we're still in the early days of AI's effect on the economy.
AI-powered writing tools are increasingly integrated into our e-mails and phones. Now a new study finds biased AI suggestions can sway users’ beliefs
www.percepta.ai/blog/can-llm...
As a research lark at Percepta, Christos embedded a computer into an LLM, showed that it could solve the hardest Sudokus, and then as a side bonus built an exponentially faster attention
🥁🥁🥁 Newly out from us today in Science Advances: “Biased AI Writing Assistants Shift Users’ Attitudes on Societal Issues”.
Large Language Models are providing users with autocomplete writing suggestions on many platforms. Could these suggestions shift users’ own attitudes? (spoiler: YES) (1/7)
"Trust becomes more valuable, not less, when the information environment degrades. The more noise enters the system, the more researchers will retreat to sources w/ human accountability. The question shifts from “what does the evidence say?” to “who is telling me this, & why should I believe them?”"
Academic publishing is currently experiencing a viral spread of “zombie citations.” I tried following one to see how these references are infecting academic knowledge systems codeactsineducation.wordpress.com/2026/01/30/t...
A Berlin court has ordered X to grant researchers API access under the Digital Services Act. Daniela Alvarado Rincón, Simone Ruf and Jürgen Bering explain how they won the case, and why it’s a major step for researcher data access.
About a year ago I made some predictions about the effect of AI on programming jobs. Block laid off 40% of its staff claiming AI made them more efficient. Is that really true or did they just over-hire? Let's look at some data and see what's really happening.
Full post: seldo.com/posts/do-ai-...
You get used to it, but it was a shock to the system at first.
“I end each day exhausted—not from the work itself, but from the managing of the work. Six worktrees open, four half-written features, two ‘quick fixes’ that spawned rabbit holes, and a growing sense that I’m losing the plot entirely.”
I got a tip that all over the world, people are using a dead-simple hack to manipulate AI behavior. It turns out changing what AI tells other people can be as easy as writing a blog post *on your own website*
I didn’t believe it, so I decided to test it myself www.bbc.com/future/artic...
I'm creating a series of short form videos about how language models work technically. The goal is to be something in between "you know it's next token prediction" and "now you've taken a machine learning class." I'd love your thoughts so here are the first few! 🧵
www.youtube.com/shorts/VZB8X...
Interactions with large language models blur the lines between what feels like human conversation and the more typical experience of using technology—which seems to be causing confusion even among what should be experienced and sophisticated users, writes Tech Policy Press fellow James Ball.
No, the US Supreme Court did not declare that AI works cannot be copyrighted
If you have been online recently you may have been a variation of this story that reads something like this: "The US Supreme Court declares that AI generated works aren't copyrightable." I won't name and shame all of the…
Interesting interview study about AI in the labor process. Best possible atmosphere (well-paid, autonomous) still results in more work, not less, because AI helps you multitask, work during nonwork time, extend your capability into other people's turf (experts then need to clean up vibejobs)
In the first essay of our new neuro-education series, @eblancosuarez.bsky.social shares how she reimagined the traditional journal club course by developing a class in which students review preprints alongside the published papers that evolved from them.
#neuroskyence
bit.ly/4reKhg8
Defuddle now has a website!
This means you can use Defuddle anywhere to get the main content of a page in Markdown format.
You can simply add "defuddle.md" before any URL, use it via curl, Skills, CLI, or add it to your app via NPM.
A group of prominent scientists launched an independent autism advisory panel over fears that Health Secretary RFK Jr. has politicized the key federal autism advisory board he oversees.
I joined 300+ security and privacy academics in calling for a global pause on online age verification: www.politico.eu/article/age-...
Text Shot: To build a useful AI model, you need to journey into the wild base model and stake out a region that is amenable to human interests: both ethically, in the sense that the model won’t abuse its users, and practically, in the sense that it will produce correct outputs more often than incorrect ones. What this means in practice is that you have to give the model a personality during post-training1. Human beings are capable of almost any action at any time. But we only take a tiny subset of those actions, because that’s the kind of people we are. I could throw my cup of coffee all over the wall right now, but I don’t, because I’m not the kind of person who needlessly makes a mess2. AI systems are the same. Claude could respond to my question with incoherent racist abuse - the base model is more than capable of those outputs - but it doesn’t, because that’s not the kind of “person” it is. In other words, human-like personalities are not imposed on AI tools as some kind of…
Giving LLMs a personality is just good engineering - Sean Goedecke www.seangoedecke.com/giving-llms-a-… #AI #training #personalities
Benj Edwards is no longer working at Ars Technica following an AI controversy that angered readers.
🚨 AI is turning research into a scientific monoculture — and as a scientist, I worry about the long-term resilience of our research ecosystem.
Our new piece in @commspsychol.nature.com with @profsanderlinden.bsky.social and Jon Roozenbeek: www.nature.com/articles/s44...
Academics and technologists are sounding the alarm about a growing crisis in scholarship as we know it: AI-generated citations of nonexistent papers that have infested real journals. Despite being fake, the sources are widely assumed to be authentic the more they appear in published literature.
"The USA And Israel Have Started Bombing Iran: A Primer
What Do We Know, What Should We Be Asking?, For What Should We Be Watching"
From @phillipspobrien.bsky.social
phillipspobrien.substack.com/p/the-usa-an...
This is one of the most reasoned & persuasive arguments for not allowing LLMs anywhere near reading & writing intensive classrooms. We can 100% choose not to outsource our reading & writing labor to a bot, and model for our students why they should do the same. #EduSky
www.tue.nl/en/our-unive...
"Powerful AI can statically help human decision-makers, but can harm collective knowledge building... it can lead to what we call “knowledge collapse” whereby in the long-run all human knowledge is ultimately destroyed.”
economics.mit.edu/sites/defaul...
