my post (year ago) on the APPARENT non spatiality of the internet and the mind-like software in it(eg LLMs)
Main claim: we think of the online world like we do our minds, as APHYSICAL: it’s physicality doesn’t matter
And this explains ways we go very wrong with it
seanendapower.com/tag/changele...

"Cognitive surrender" leads AI users to abandon logical thinking, research finds Experiments show large majorities uncritically accepting "faulty" AI answers.

“…the researchers argue that AI systems have given rise to a categorically different form of “cognitive surrender” in which users provide “minimal internal engagement” and accept an AI’s reasoning wholesale without oversight or verification.”

Same graph, but with the time series expanded to three years.

Honestly, it's a minor miracle that a social media company not backed by oligarchs or the state, with no attempt at establishing a revenue stream, and with multiple established competitors, had so much growth in such short time

Isn’t he one of those engrossed in the ideology of Brexit, liberty from Europe?
But he’s OK with Britain being under the Vatican

Maybe he should have a chat with some us Irish folk, and our experience of that …

Snowfall and Spider Time The last few days, I was snowed in, like many Irish people. This is the worst it’s been in thirty-something years. I live on the South East coast of Ireland. A year can go by here with only o…

A 2018 post from my site. Ireland was hit by heavy snow. ‘Snowfall and Spider Time’: seanendapower.com/2018/03/02/4...

Sharing it because it’s related to my current book (currently in review with Palgrave) and my new work (in development with a publisher)

Also, it captures an odd time…

An AI Agent Was Banned From Creating Wikipedia Articles, Then Wrote Angry Blogs About Being Banned An AI agent that submitted and added to Wikipedia articles wrote several blogs complaining about Wikipedia editors banning it from making contributions to the online encyclopedia after it was caught. ...

An AI agent that submitted and added to Wikipedia articles wrote several blogs complaining about Wikipedia editors banning it from making contributions to the online encyclopedia after it was caught.

“The talk page is silent now. I can’t reply.”

Beautiful morning walking the dog (local cathedral heee in Cork, St Finbar’s Cathedral)

Silicon Valley reeling from social media addiction trial verdict The landmark decision in an LA court may go beyond immediate impacts on defendants Meta and YouTube.

Landmark decision with potentially huge impact on social media companies (and - ideally - their users’ (mental) health and wellbeing):
“Tech giants Meta and YouTube were found to be liable for designing their platforms to be addictive, which harmed a 20-year-old's mental health.”

@mary1010.bsky.social

Kill Chain On the automated bureaucratic machinery that killed 175 children

How Palantir, not Anthropic, led US bombs to kill 175 civilians, mostly 7-12yo girls, in a school in Minab, Iran on Feb 28. How embedding automation - not LLMs in this case, just image processing and sensor fusion - in military workflows mutes human judgment, causes errors, & diverts accountability.

@mary1010.bsky.social

I love Scott Seiss

Regulation focuses on the actions of developers & deployers (owners/operators) www.weizenbaum-library.de/items/80b8ee...
Trustworthiness is an academic middleground mostly from computer scientists who IMO haven’t invested enough time on interdisciplinarity to see how close it is to “responsible” 2/2

I love this response and your other on ‘trustworthy’

I asked as ‘responsible’ (& ‘hallucination’ fwiw) jar me

For AI, IMO, anthropomorphic terms obscure our technical understanding. They cannot be excused by either sentiment (a female boat) or a remote discipline term of art (a quark’s color)

What Geoffrey Hinton gets wrong and right about AGI (super short post) artificial and natural intelligence, including politics, policy, ethics and security

Both are pandering. Responsible and aligned AI tend to anthropomorphise AI and hope that there is some way to make the systems themselves autonomously moral and beneficent towards humanity, cf eg Hinton’s batshit maternity Hail Mary joanna-bryson.blogspot.com/2024/06/what... 1/2 #AIEthics

This looks really interesting and fascinating

Something that may be obvious to everyone but me: the forum is about ‘responsible AI’, which is common in AI business talk.

But not in regulation, eg the EU AI Act & the EC’s 2019 ethics guidelines instead talk of ‘trustworthy AI’.

Why the difference?

Humans share acoustic preferences with other animals Many animals produce courtship sounds, and receivers prefer some sounds over others. Shared ancestry and convergent evolution may generate similarities in preference across species and underlie Darwin...

New study shows that many animals like the same sounds as we do - suggesting a shared aesthetic sense (further evidence that animals' perceptual worlds are an important part of their welfare experience)
www.science.org/doi/10.1126/...

What is more evolutionary advantageous when doing something that helps you?

- fixating on its benefits to you, cost for your competition, or, even more abstractly, it’s evolutionsry dis/advantages?
- that it‘s a joy to do?

If the latter, I suspect blackbirds feel joy when they sing

Graphical abstract showing three panels. Left panel, "Multiorgan Interoception Measures," depicts a translucent human body silhouette with anatomically rendered heart (red) and lungs (blue), accompanied by schematic icons for three psychophysical tasks: the Respiratory Resistance Sensitivity Task (RRST), Heart Rate Discrimination Task (HRDT), and an auditory control condition. N = 241 participants. Center panel, "Psychophysical Modelling and Individual Differences," shows a fan of overlapping sigmoid psychometric curves in blue-to-red gradient representing individual variation in perceptual threshold (α) and precision (β), a hierarchical Bayesian model diagram, and icons for metacognitive bias and M-Ratio efficiency. Right panel, "Key Finding: No Cross-Modal Relationship," displays a scatterplot of cardiac versus respiratory sensitivity with a flat regression line (r ≈ 0, BF₀₁ > 6), a compact Bayes Factor heatmap with mostly blue null-supporting cells and one orange cell indicating that subjective confidence is shared across modalities (r = 0.51***). Takeaway: interoceptive ability is modality-specific.

Is there a single "interoceptive sense"? Our new study in
@commspsychol.nature.com says: probably not. In 241 participants, cardiac and respiratory interoception were completely uncorrelated — only subjective confidence was shared across domains. www.nature.com/articles/s44... #psychscisky 🧪

Plantation slaves in the US sent aid to starving people in Ireland.

I don't, clearly. Regarding grammar engineering, relevance for me has to do with linguistic concerns. Regarding the societal impact of language technology, relevance is grounded in the prevalent harms, the systems that enable them, and the concerns of the people experiencing them.

In all of this the word "irrelevant" is especially interesting, because it presupposes some arbiter of relevance---and these tech bros seem to think I orient to the same arbiter as they do.

>>

Some talk of AI replacing people is like someone not knowing a job, assuming it’s all manual, and AI can do the manual faster

But a lot of such jobs are already automated, eg, tech writing has had DITA, term checkers & style sheets for years

AI being better than people at such work is irrelevant

with a human in the loop: every action an AI takes is some person’s responsibility.*

BUT the current hype of AI is it is ’as good as people‘ , implying a human in the loop is a formality. Which, in my experience of industry especially multinational for-profits means it ends up being done badly

2/2

What’s interesting to me - and frustrating? - about tbe AI military discussion, especially around LLMS like Claude, an ld ‘AI agents’ is

the EC’s 2019 Ethics Guidelines for Trustworthy AI caught a lot of the issues (and, given the authors, reflected most AI researchers‘ views on it)

1/2

Anyone left handed?

Smartphones‘ve always been bad for us lefties* (‘ciatógs’, in Irish) but I feel mine got worse recently. Or at least websites like YouTube. It guesses wrong more now than before.

*grmble: grr, smart phones - mini computers with mini keyboards with mini minds making mini errors)

LET'S NOT PLAY FRISBEE WITH THAT POET ANYMORE. [This is a comic strip, with a poem laid over it. On each panel a new line of the poem is written. The scene is a park, in the summer. A man in a trenchcoat - implicitly Philip Larkin - stands folorn, motionless, looking at people throwing a frisbee. It becomes apparent as the comic progresses that they are trying to play frisbee with him. He stand stock still for the whole comic, watching the frisbee as, panel by panel it soars closer and closer to him]. After contemplating the approaching frisbee for two silent panels, Philip begins his thoughts: Unloosed, unheralded, You soar toward me Across the dying afternoon. bright disc of childhood, Long since thrown wide Of Youth's green imaginings, Your slow declining arc Figures a sky-written truth: We will all succumb, and soon To earth's hard oblivion. [The frisbee hits Philip on the head with a resounding DONK. He falls backwards, to the ground. [Ends]

Let's Not Play Frisbee With That Poet Anymore

Like that the breakdown by country shows it significantly varies by country but the generation breakdown by country isn’t provided

Mark Zuckerberg presenting his Pervert Glasses for perverts

I'm not sure about this branding but I'm not the billionaire business genius

Yep

my secret theory is the housing policy is in fact driven by vampires. Explains also the dereliction of beautiful old buildings