Record-setting temperatures and rainfall in the Arctic over the past year sped up the melting of permafrost and washed toxic minerals into more than 200 rivers across northern Alaska, threatening vital salmon runs, according to a new report. nyti.ms/48Qr26R

"LLMs can be accurate", proponents argue, they just need a bit of oversight. And that oversight doesn't have to mean human-in-the-loop. It can come from other LLMs." Well, clearly not.

Yet one more reason we cannot allow LLMs to serve as epistemic grounding is that we cannot triangulate among them the way you can among reasonable independent sources. They bullshit in the same way and end up agreeing with one other about things that are completely false.

There's a cross-platform uniformity to how LLMs bullshit you. Today I was designing an exercise for a responisble-use of-LLMs course we are designing for NIH trainees. I asked ChatGPT to complete one of Claude Shannon's original guessing games: "There is no reverse on a __(10 letters)__"

🌱 Join us Wed Dec 10 @ 9 AM, l’École normale supérieure (Paris) to discuss complex questions on AI’s ecological impact & sustainability with Marlène De Bank, Gauthier Roussilhe, Sasha Luccioni & more. Schedule and inscriptions: www.ens.psl.eu/agenda/works... @oii.ox.ac.uk

Honoured to be nominated Young Academic by the @academiajoven.bsky.social! These past 2 days in Valencia have been enriching, surrounded by colleagues with incredible achievements. Over the next 5 years, I'll focus on exploring how AI is shaping science&society. Excited for the journey ahead!

Palantir CEO Says Making War Crimes Constitutional Would Be Good for Business

¿Cómo se puede hacer una estrategia nacional de salud e IA sólo citando estudios de empresas que son beneficiarias de comprar sus productos?¿cómo puede ser que se hagan tales afirmaciones sin ningún tipo de comprobación? cc @javierpadillab.bsky.social @htejero.bsky.social

'Oxford’s governance is under increased scrutiny following a Bloomberg investigation, published on 19 November, which reported claims the university has been slow to act against male academics accused of inappropriate behaviour.' 1/2

V Jornada de la @academiajoven.bsky.social "La Ciencia frente a las grandes Crisis" 🗓️28 noviembre ⌚️9:00-18:00 horas ➡️Presencial y online 📍Jardín Botánico de la Universidad de Valencia Más información e inscripción: academiajoven.es/proximos/jaj...

Just published in open access: Early Modern Narratives of Islam across Europe: A Connected Memory: brill.com/display/titl...

This is what happens when STEM folks don’t take humanities. One anthro course covering the Sapir–Whorf hypothesis would have cleared this up early.

I don't know about you but the way my brain works is by analyzing the contents of the entire internet to make an educated guess about what word I should use next.

I’ve been running around asking tech execs and academics if language was the same as intelligence for over a year now - and, well, it isn’t. @benjaminjriley.bsky.social explains how the bubble is built on ignoring cutting-edge research into the science of thought www.theverge.com/ai-artificia...

Much like offloading onto LLMs, men are more likely to feel OK offloading generally. Women do not & sadly are used to doing this work. Men are formed by patriarchy to demand, expect, and tolerate this. This is in pure contrast to all the women who contact me to say they will work hard to fight AI.

Researchers attending include DPhil students Andrew Bean, Ryan Brown, Franziska Sofia Hafner, @ryanothnielkearns.bsky.social , Harry Mayne and Kaivalya Rawal; Shreyanash Padarha, Research Assistant and faculty members @computermacgyver.bsky.social Adam Mahdi, @rocher.lc and Chris Russell. 2/2

"What a country calls its vital economic interests are not the things which enable its citizens to live, but the things which enable it to make war." Simone Weil ( 1909–43).

I don’t want to seem out of touch but I don’t actually understand the economy anymore.

Loved this bit : “The most lucrative users – English-speaking professionals willing to pay $20-200 monthly for premium AI subscriptions – become the implicit template for ‘superintelligence’.”

Join us online this Thursday!

🤖 "Holes in the web: Huge swathes of human knowledge are missing from the internet. By definition, GenAI is shockingly ignorant too." Excellent essay on algorithmic epistemological knowledge and the collapse of knowledge throughout mean-driven data machines. aeon.co/essays/gener...

Public Data Cultures is out today on @politybooks.bsky.social! 📗🎊 jonathangray.org/2025/10/23/p... It aims to nurture critical and creative engagements with public data as cultural material, medium of participation and as site of transnational politics. Happy to see it out in the world! 🦔🌌💜

🌐🚨 CALL FOR PAPERS 🚨🌐 'The Visual Politics of Digital Ecologies' 🗓️ 2-3 February 2026 📍 @oxfordgeography.bsky.social We invite one-word titles (e.g. hallucinating, zooming, obscuring, glitching, generating) Deadline: 14th November Full info: digicologies.com/cfp

📌 Utrecht University's CDH online lecture: "Following the thing AI – from ecological ruinations to ecofeminist imaginations" 👉Thursday 20/11 at 15:30 cdh.uu.nl/event/cdh-on...

This is going to be a big old thread where I dump thoughts, reactions, charts etc from the @iea.org World Energy Outlook 2025 :) www.youtube.com/watch?v=wlkP... www.iea.org/commentaries...

“A recent response from the European Commission to our access to documents request not only revealed that the President of the Commission was repeating the words of tech CEOs, but it also accidentally divulged the Commission’s use of generative AI in public documents.” 🤦‍♂️

"People have to have a language to speak about what other possible futures are available to them,” Hall wrote. One task...is therefore to devise a language that expands the sense of what is possible" NEW: Casey A. Williams reads Stuart Hall for the Climate Crisis www.break-down.org/reading-stua...

Fantastic interview with @anavaldi.bsky.social pointing out the risks of facial recognition. But, lo and behold, the caption is very misleading. Gracias Ana por hablar sobre este tema tan importante para quienes migramos a UK.

Really good, short, LRB blog by @michaelchessum.bsky.social abt the rise of the Greens... Spoiler - his conclusion - "For more than a hundred years, the Labour Party has had a monopoly on political representation on the British left. That is now breaking down." www.lrb.co.uk/blog/2025/oc...

Matthew Gandy (@cam.ac.uk) will present on “Attentive Observation: Walking, Listening, Staying Put” tonight at 18:00 as part of the interdisciplinary lecture series “On Environment.” Find out more: www.carsoncenter.uni-muenchen.de/events_conf_... #lectureseries #rccevents #envhum #envhist

Our paper on clusters of schools across the UK is out. We find five clusters: multi-racial & super-diverse middle & working-class schools of towns & cities, suburban white middle class schools, established elite schools & (post-)industrial white working class schools. doi.org/10.1002/berj...

Wim Vanderbauwhede: “You give an input and it ‘guesses’ the answer a user would want to get, based on the input and the data it was trained on. The ‘machines’ don’t think; they generate something plausible, something that seems acceptable.” apache.be/2025/10/24/b...

Who today is influencing science a la Big Tobacco or Oil and Gas? In our new preprint we show how tech companies like Meta are capturing research on their product, using mechanisms that subtly (or not so subtly) shape what science gets produced

It's perhaps pertinent to note that people in reply *don't believe* these numbers. Yet they are from a UKRI spreadsheet. So is UKRI wrong? Or are the numbers just unbelievably bad?

Last time I checked those were at roughly 5% success rate as per the AHRC website.

But column two says "Year application received". This to me suggests that the success rate is "the probability that you would have been successful by Oct. 2025, given that you submitted in 2025", which is different from "the probability that you will be successful, given that you submitted in 2025"

This is at a time when universities are turning more to grant capture as the primary currency for promotion etc (usually regardless of what you do with it)

CAIL = prerequisite knowledge for a critical perspective, such as to tell apart nonsense hype from true theoretical computer scientific claims. For example, the idea that human-like systems are a sensible or possible goal is the result of circular reasoning and anthropomorphism. olivia.science/ai 🧵

Makes the ERC Starting Grants look very attainable in comparison (12.2% success rate for latest round)

My performance (and employability, let alone promotability) is based on my ability to win these grants. Even if i did nothing but wrote them (+ all the networking & other prep work) I could easily go years - decades - without winning one, according to these stats.

1% for new investigator is particular obnoxious as it would seem to directly undermine the core aims of the scheme. You're not helping ECRs develop if you won't give any of them money to manage as PIs

When a Microsoft data center opened next to the small Mexican town of La Esperanza, locals say electricity outages and water shortages grew severe. The local doctor rushed patients a half hour to the nearest hospital when blackouts cut off his oxygen concentrator. www.nytimes.com/2025/10/20/t...

📺 TV en DIRECTO | Sánchez pedirá a los Veintisiete fondos para viviendas protegidas y un mecanismo europeo para frenar la especulación: "Es la primera vez que el Consejo Europeo se va a hablar de la emergencia habitacional que vive toda Europa" tinyurl.com/2ad8we5s

Thought provoking article by @nytimes.com highlighting the hidden cost of AI, its environmental impact and the consequences for local communities with commentary from @anavaldi.bsky.social @oii.ox.ac.uk.

Can AI simulations of human research participants advance cognitive science? In @cp-trendscognsci.bsky.social, @lmesseri.bsky.social & I analyze this vision. We show how “AI Surrogates” entrench practices that limit the generalizability of cognitive science while aspiring to do the opposite. 1/

Great article. The point is that, even if the sample is unrepresentative, it is larger at the bottom. This will impress reviewers, journalists and, of course, the university's department of communications. Also, AI is a tool that allows bypassing Ethics committes and thus speeds up research.

Wow. Harvard nuking its PhD programs - Science PhD admissions reduced by more than 75% - Arts & Humanities reduced by about 60% - Social Sciences by 50–70% - History by 60% - Biology by 75% - The German department will lose all PhD seats - Sociology from six PhD students to zero

After submitting a FOIA request UKRI, I obtained success rates by three grant call scheme and I can only say that I am disheartened by the results: - AHRC Responsive Mode 2025: 2% - ESRC New Investigator Grant 2025: 1% - ESRC Research Grant Round 2025: 1%

Very glad to see this piece by the @nytimes.com on data center struggles from Mexico to Ireland: "Government support worldwide has helped tech firms build with little accountability, said Ana Valdivia, an Oxford University lecturer studying data center development." www.nytimes.com/2025/10/20/t...