We knew this all along, but ok

How strange that LLMs can't do a thing they were never even designed to do in the first place

To me, this is theoretical (=scientific) *psychology* at its best. Thanks to these authors, we now know some new, fairly universal (abstract), undeniable (proved), and useful facts about *affordances*.

We posit that the relationships of transformations of the world resulting from the agent's actions should be included in the agent's representations of the world, 2/n www.sciencedirect.com/science/arti...

Ablation study and resulting optimal architecture that considerably improves recent research on CNN-Hebbian learning integration with competition mechanisms. with @julian-jn.bsky.social www.sciencedirect.com/science/arti...

🚨Our preprint is online!🚨 www.biorxiv.org/content/10.1... How do #dopamine neurons perform the key calculations in reinforcement #learning? Read on to find out more! 🧵

Move info on this website: olivia.science/ai by @olivia.science QR code goes here: doi.org/10.5281/zeno...

Just think of all those Generative Boris Johnsons heading into the world in a few years...

This might be useful too especially the proof for such systems bsky.app/profile/iris...

People lost the North long ago. Now they have lost themselves in oblivion.

As soon as people begin to separate others into with or against us, they start to lose the large majority of ppl who are neither and who could otherwise be attracted with kindness and an effort to explain, and also those who utterly dislike being dragged towards any faction or monocratic stance.

on a dire turn of events, reverting from online to in-class exams led to a 20% decrease in class average for the first exam. idk how much of it is due to the absence of chatGPT access but i suspect a lot. especially since it's an open-notes test, and time pressure wasn't at all an issue.

Why is it that when you check some numbers and note them, they almost immediately change? Let's call it "the quantic curse of the curious" 🤣

My last contribution to the programme: sustaining the high proportion of women, at around 28% so far this year.

Mushenge

Ex nihilo nihil fit, but AHA"s nihilism may be rewarded

Too many education bureaucrats and administrators are in the pocket of corporate interests. The deeply corrupt laziness of policy makers who push AI into our schools must be resisted by everyone who cares about quality education. pivot-to-ai.com/2025/09/13/c...

We need something like this here today.

Brilliant! Where is the malicious behaviour of projecting publicly available pictures? www.bbc.co.uk/news/article...

Freaking anti-vaxxers collapsing one of the main arteries in London on the day the pumpkin king is visiting. Tsk, tsk Where are the police now?

Jeez! Predatory journals are getting increasingly audacious! 🫤

And we are going to do what about it?

All recent years' efforts have gone down the drain. I am used to it; I am old and resilient: Stop moaning, back to work, it's just a matter of starting all over. Right? Now, have these luminaries spared some thought about how this will affect the youngsters? Clearly not.

>> and that features that cause transformations that do not form group structures, yet show partial and local symmetries, are also necessary to build "good" representations and powerful learning mechanisms. 3/3 www.sciencedirect.com/science/arti...

Quoting a good friend: "They treat people as tools, 'useful idiots' even". True, but only because we let them deceive themselves into being superior. They need to believe it to keep breathing.

I lived the times too, all too familiar. I'd like to encourage younger people to read this, especially men who feel somehow oppressed by the "woke" culture. www.theguardian.com/commentisfre...

What an idea! Shall we apply it and swap university executive teams with AIs? Why not! 😶 🙄 www.bbc.co.uk/news/article...

Heads are not equal to leaders. If only humans would know! 🙃 "But there are other aspects of power: the idea of leadership — for example, who determines where the group goes today or whether to fight against a neighboring group — and these aren't necessarily linked to the same hierarchies."

A quick summary for busy people: We propose a mathematical framework for learning representations by extracting the algebra of transformations of worlds from the agent's perspective. 1/n www.sciencedirect.com/science/arti...

Hmm... so universities can't afford "unfunded amateur research," but they can afford to pay exorbitant amounts to arrogant and incompetent businessmen.

Hmm... so universities can't afford "unfunded amateur research," but they can afford to pay exorbitant amounts to arrogant and incompetent businessmen.

"We argue against treating reproducibility as an inherently desirable property of scientific results, and in favor of viewing it as a tool to measure distance between an original study and its replications." Now working with collabs to nail down "replication of what" on genomics study? Exactly.

I enjoyed this interview very much, which says something coming from someone who doesn't fancy them! 🎯"antithetical to the whole enterprise of academia and science" "Prototypical pseudo-scientific strategy." "We don't read any more" "uniqueness is filtered out" "regurgitating the past" 🎯💣

My talk today, live streamed at 3pm UK time, 10am EST from @memoryplace.bsky.social You need to registered to get the link, see below:

As promised, scrap about paranoia and its underlying mechanisms continues: osf.io/preprints/ps...

A quick summary for busy people: We propose a mathematical framework for learning representations by extracting the algebra of transformations of worlds from the agent's perspective. 1/n www.sciencedirect.com/science/arti...

One "dear pal" down, many still to go. Especially one! www.theguardian.com/politics/202...

We have a problem with A. It generates an incredible administrative burden. Solution: revise A by creating 3 or 4 concomitant processes (which may or may not be exclusive). Now we no longer have an A problem, we have A-ish, B, C, D and E.

Sadly, it is increasingly difficult to believe anything related to science that has the support of the American president. Anything, in fact. Liars, lie. nasawatch.com/astrobiology...

So we have a government and a police body that hurry to arrest peaceful eighty-somethings demonstrators, and this rascal is still free? 😡 www.theguardian.com/uk-news/2025...

Enjoying this quote too much. It has often been attributed on social media to Mario Benedetti: "When a mediocre person is given a gram of power, they think they have a ton of authority." (The translation is mine.)

Can you all explain yourselves what the actual f is 23k views and about half as many downloads on a preprint right now come off it 🥺

‘Microsoft makes a lot of money from this partnership, it is simply a supplier of defence to Israel and other countries,’ says @djoerd.idf.social.ap.brid.gy 👏🏼 www.voxweb.nl/en/professor...

I am attending an event on genAI and education. The word is "hysterical" views. It has been repeated by each speaker so far. I may need to leave for my mental health. 😡

We certainly do not live in the same country/world we lived in a few years ago. This is an embarrassment for us all. Not only are they supporting the most cruel war and a genocide, but they also want to force us to be complicit. >> youtu.be/coPyDlZL1l4

Where is the so famous British punctuality? Late for an announced test alarm! Tsk, tsk 🤓

Mood.

I deeply agree with the sentiment of this paper. Although, as a norm, I dislike mixing up "activism" and research, there are occasions in which it is justified. Universities and society are blindly and impositively adopting corporative tools that severely jeopardise our capabilities of reasoning >>