Screenshot of https://www.complexdatalab.com/stamina/ web page

Lab knowledge base on Semble spotted in the wild 👀
@stamina-workgroup.bsky.social

Open-mindedness was the strongest predictor of *rejecting* conspiracy theories in a sample of 46,745 participants in 68 nations

In particular, participants who were threatened by people who disagree with them were the most likely to believe conspiracy theories.
www.cambridge.org/core/journal...

We are very much looking forward to it!

I'm not an expert here, but my impression is that degrowth did provide a tent in which to do meaningful analyses to debunk a neoliberal end-of-history. It just never got much power to do things. Maybe decomputing could play out differently. more precarity could mean more appetite for change? I dunno

Very fashy interior design. I guess it's a good thing that it seems out of place enough to me to come off as part of the alternative timeline segment of some sci-fi show.

you can see in a near future of even more precarious living with AI labeled as the cause, how "decomputing" as a reactionary humanist/luddite movement to technologies of control (but really just overall malaise) could capture minds as much or more than degrowth ever did. Will it be productive? 🤷‍♂️

The Guardian view on defence spending: should the UK’s security rest with Donald Trump? | Editorial Editorial: A former Nato chief demands more cash while fixing Britain’s global role. Before billions are spent, ministers must define the purpose of its military

“As Khem Rogaly argues, defence spending provides a weak economic stimulus compared with public investment – and is even worse as a job creator.”

Khem Rogaly quoted in @theguardian.com ’s editorial on military spending 👇

www.theguardian.com/commentisfre...

Why interrogate bias as social (e.g. in dilemmas), when bias is already there for LLMs sampling from simple discrete and continuous distributions, and even just the notion of LLM producing calibrated responses has issues e.g. a paper from only a few days ago: arxiv.org/pdf/2604.065...?

What moral framework would you argue from to defend your claim that uniform selection here is considered the fair rule?

🚨Join us for our next www.complexdatalab.com/stamina/ talk @ 11am ET tmr!

@zhijingjin.bsky.social will talk all about LLM cooperation!

Standard social annotation apps include mainly tagging, bookmarking or highlighting - super excited to expand the sensemaking toolbox with Connections!

Let's make the semantic web great again ;)

*capability, not capacity

Nicholas Carlini - Black-hat LLMs | [un]prompted 2026 YouTube video by unprompted

The concern precedes the press release. e.g. longtime security egghead & recent Anthropic employee Carlini visibly shaken by the rate of increase in capacity for finding deep vulnerabilities and saying the next 2 months is a crucial window to do something www.youtube.com/watch?v=1sd2...

Feels so wrong that price I pay at the pump will respond to how effective the bombing campaign is.

"AI is an incredible force multiplier for implementation, but it’s a dangerous substitute for design"

Improving Latent Generalization Using Test-time Compute Language Models (LMs) exhibit two distinct mechanisms for knowledge acquisition: in-weights learning (i.e., encoding information within the model weights) and in-context learning (ICL). Although these...

When and how can test-time thinking allow models to use information latent in their training data? What are the benefits and tradeoffs relative to other solutions like synthetic data augmentation? Pleased to share (after a long delay) an exploration of these issues: arxiv.org/abs/2604.01430 thread:

I like how Anthropic is just obliquely releasing their work on recursive self-improvement.

Trump's hackin' the mainframe with vibe-coded press conferences, lol.

Just realized Trump is quite literally "working the stock market" insofar as he is operating a kind of information engine that extracts work ($) from Trump-announcement-based market "corrections".

I feel like the development wave in AI agents now to have very liberal memory access is going to blowup in their faces. There will be a course correct that introduces a more conservative standard that is much more cautious about how correlations are managed across sessions/context.

In Silico Sociology: Forecasting COVID-19 Polarization with Large Language Models By training deep neural networks on massive archives of digitized text, large language models (LLMs) learn the complex linguistic patterns that constitute historic and contemporary discourses. We argu...

This wild paper showed that GPT-3 (whose cutoff is pre-pandemic) seems to encode the kind of social pola
rization we observed in the pandemic, suggesting the potential for polarization was baked into the culture filter through its weights. arxiv.org/abs/2407.11190

Strange loop. Like a liar's paradox, but in the mind of the liar.

freely available after the conference I presume. Let us know :)

Can I read about Lea?

@spacecowboy17.bsky.social is there a clear culprit here in the For You knob set?

Another “atproto magic moment” (tm): @semble.so bookmarks alongside @margin.at highlights. Oh and hey that’s the @co.cameron.stream bot over there making annotations. Easy to take this stuff for granted like it always existed but it’s crazy cool and kinda mind blowing when you zoom out a little

It's in 5 billion, not 500 million, no? 10x more. Somewhat less pressing.

We need menswear guy to comment on the outfit.

great paper to point to now when referencing this top-down influence. I think finding out how to tease its effects apart from those of the bottom-up agent planning influence will lead to real advances in understanding and modelling social systems.

New paper: “𝐀 𝐓𝐡𝐞𝐨𝐫𝐲 𝐨𝐟 𝐀𝐩𝐩𝐫𝐨𝐩𝐫𝐢𝐚𝐭𝐞𝐧𝐞𝐬𝐬 𝐓𝐡𝐚𝐭 𝐀𝐜𝐜𝐨𝐮𝐧𝐭𝐬 𝐟𝐨𝐫 𝐍𝐨𝐫𝐦𝐬 𝐨𝐟 𝐑𝐚𝐭𝐢𝐨𝐧𝐚𝐥𝐢𝐭𝐲”

Agent-based models of social order work better when agents act by predictive pattern completion from prefix (culture/context) to suffix (action) than when they act through expected value maximization