New paper alert🚨
www.sciencedirect.com/science/arti...
Two takeaways. First, 2- to 4-year-olds can use a process called second-order correlation (SOC) learning -- e.g., A & B go together and B & C go together, then A and C go together -- in a category context to make causal inferences.
🚨New Paper!🚨 How do reasoning LLMs handle inferences that have no deterministic answer? We find that they diverge from humans in some significant ways, and fail to reflect human uncertainty… 🧵(1/10)
04.03.2026 16:13 — 👍 50 🔁 18 💬 3 📌 1
⚡ Next Monday!
Historical & Philosophical Foundations of Artificial Intelligence Conference
Speakers:
Tim Crane (CEU) @timcrane.bsky.social
Thomas Haigh (UW-Milwaukee)
Matthew L. Jones (Princeton)
Raphaël Millière (Oxford)
Amira Moeding (Cambridge)
Shannon Vallor (Edinburgh)
Tickets 👇️
#philsky
Transformers by Raphaël Millière: https://doi.org/10.21428/e2759450.d3acfbfb
03.03.2026 12:00 — 👍 6 🔁 4 💬 0 📌 0
New and free online: @msgjonhere.bsky.social
& my edited collection of essays on belief with Oxford University Press:
academic.oup.com/book/62410
Table of Contents in thread
Interviewed for @joeykpollock.bsky.social’s blog. Made to account for my errors in public.
01.03.2026 09:27 — 👍 23 🔁 4 💬 1 📌 2
Every sentence in this post is a masterpiece of anthropomorphic absurdity:
substack.com/home/post/p-...
Callers to Washington state’s driver’s license agency who select automated service in Spanish are instead hearing an AI voice speaking English with a strong Spanish accent.
27.02.2026 17:30 — 👍 1988 🔁 733 💬 133 📌 513
In one four-day period in March 2025, engineers logged 205 maintenance calls. Sailors reported working up to 19-hour days to fix leaks and restore suction, with problems in a single head capable of disabling an entire section of the ship. Since 2023, the Navy has carried out at least 10 acid flushes, each costing about $400,000, though the procedure cannot be performed while the carrier is underway.
The sailors on the USS Gerald Ford who keep stuffing their t-shirts into the toilets so the bathrooms flood and the carrier can’t deploy to Iran should be awarded a collective Nobel Peace Prize
27.02.2026 11:35 — 👍 2905 🔁 737 💬 40 📌 61
NEW POLL on Dem. structural reforms: U.S. adults support 18-yr term limits for Supreme Court justices by a 50-point margin (GOP is +34 in favor), favor statehood for Puerto Rico and limits on pardons by POTUS, and are split on DC statehood, packing the Court:
www.gelliottmorris.com/p/two-thirds...
The party that gets aggressively on the right side of this issue is going to win a lot of elections in a lot of places over the next decade. www.nbcnews.com/news/us-news...
26.02.2026 13:08 — 👍 2207 🔁 790 💬 115 📌 42
excited to share some recent work!
neural networks trained on multi-view sensory data are the first to match human-level 3D shape perception
we predict human accuracy, error patterns, and reaction time—all zero-shot, no training on experimental data
arxiv.org/abs/2602.17650
1/🧠
🚨New preprint! In-context learning underlies LLMs’ real-world utility, but what are its limits? Can LLMs learn completely novel representations in-context and flexibly deploy them to solve tasks? In other words, can LLMs construct an in-context world model? Let’s see! 👀
26.02.2026 17:28 — 👍 37 🔁 5 💬 1 📌 1
I enjoyed talking with @kensycoop.bsky.social on the Many Minds podcast about the metaphors we use to conceptualize AI.
manyminds.libsyn.com/seven-metaph...
This is the swamp in which we work. To pretend that these men are exceptional mistakes the deep misogyny & racism built into American higher ed. These same men have fashioned themselves as heterodox thinkers against the world of woke but this is who they really are. www.chronicle.com/article/unma...
26.02.2026 11:29 — 👍 165 🔁 63 💬 3 📌 2
In a forthcoming paper we found little evidence that those persuaded by LLM-generated moral advice were evaluating the reasons in the advice in any meaningful way.
Nor were they less persuaded by an LLM that had previously produced obviously flawed outputs
philarchive.org/rec/LANPDT-2
AIs can’t stop recommending nuclear strikes in war game simulations
Leading AIs from OpenAI, Anthropic and Google opted to use nuclear weapons in simulated war games in 95 per cent of cases
www.newscientist.com/article/2516...
Functional MRI scans of more than 100 2-month-old infants suggest that they are capable of distinguishing among a variety of different objects. The findings challenge perceptions of cognitive development as a gradual process.
By @helenak.bsky.social
www.thetransmitter.org/cognitive-ne...
A new preprint, co-authored with @johnwkrakauer.bsky.social:
The Deliberation Taboo
Cognitive science is, nominally, the science of thinking. We argue that the field has no theory of what thinking is and, even worse, that the topic has largely dropped out of focus. 1/
osf.io/preprints/ps...
How can it be that modern LLMs are so bad at playing games? Aren't they supposed to be generally intelligent? Honestly, they are better at coding games than playing them. Maybe programming is just a particular type of game? Our new position paper tackles these questions.
24.02.2026 05:41 — 👍 18 🔁 2 💬 1 📌 1The Epstein files document what many women researchers have long experienced but rarely seen laid bare so starkly: exclusion operating behind closed doors, shaping who gets funded, invited, mentored, and taken seriously. How many of these networks, norms, and gatekeepers remain in place?
23.02.2026 23:35 — 👍 4284 🔁 1767 💬 42 📌 52
Today we present a new framework for measuring human-like general intelligence in machines: studying how and how well they play and learn to play all conceivable human games compared to humans. We then propose the AI Gamestore a way to sample from popular human games to evaluate AI models.
23.02.2026 15:49 — 👍 20 🔁 7 💬 1 📌 0
5. The US economy’s dependence on AI may have been overstated, experts say. In 2025: it became conventional wisdom that AI was a major driver of growth. But some prominent economists estimate that it contributed as little as 0% to economic status
Aren’t we lucky that universities haven’t gotten carried away with the hype and taken a careful approach to AI?
Imagine if they had decided to push AI into our teaching and work life!
All praise to our leaders in higher ed for not falling for hype, hallelujah and praise the goddess
“Cognitive psychology has shown that students grow intellectually through doing the work of drafting, revising, failing, trying again, grappling with confusion, and revising weak arguments. This is the work of learning how to learn”
phys.org/news/2026-02...
🚨🚨New Preprint Alert!🚨🚨
www.biorxiv.org/content/10.6...
Animal learning is painfully slow (at least initially). Yet, well trained animals can learn very fast, sometimes displaying few-shot inference. How does this transition occur?
1/5
Edward Deci (1942-2026) started it all. Still in his 20s, he found that people who were rewarded for doing a puzzle lost interest in playing with it later (vs. people who hadn't been rewarded). Turns out motivation isn't a single thing. And the extrinsic kind can undermine the intrinsic kind.
Absolutely wild. WSJ reporting that some OpenAI staffers thought the Tumbler Ridge shooter's use of ChatGPT raised the potential of real world violence but leadership decided not to alert RCMP. www.wsj.com/us-news/law/...
20.02.2026 23:28 — 👍 1701 🔁 837 💬 33 📌 140
