who has the upper hand in the Iran war according to the former MI6 chief? @shashj.bsky.social @economist.com

www.economist.com/insider/insi...

funniest breaking news ever😂 @financialtimes.com

It's not Roosevelt we're dealing with. Hell, it's not even Nixon.

😞

what Rousseau was made to believe early in life: “Despite what my exterior appearance and animated features might seem to promise, I was, if not absolutely inept, at any rate a boy of small intelligence, lacking in ideas, practically without accomplishments, in a word very limited in all respects.”

great success🎉

contemporary rephrasing:

this is the way the world ends
not with an epic fury
but with an epic failure

Anthropic to sue Trump administration after AI lab is labelled security risk Pentagon has banned the start-up from government contracts in feud over military use of its technology

my subscription to Anthropic’s Claude has expired exactly today - perfect timing to renew it👌🏻

economist.com/obituary/202...
from The Economist

3 philosophical debates from the 20th century that neuroscience is reshaping Modern neuroscience is reshaping how we understand free will, meaning, and the self by revealing them as emergent features of the brain.

"For all practical purposes, brains and behavior must be described probabilistically." @drrachelbarr.bsky.social

It could also mean that LLMs struggle with 𝘢𝘣𝘥𝘶𝘤𝘵𝘪𝘰𝘯, i.e. inferring the best explanation from incomplete evidence, particularly when that evidence consists of weak signals, anomalies or emerging phenomena that do not yet form a clear pattern.

What the practitioners seem to be describing is something more specific: a problem with the kind of inductive generalisation LLMs perform - one that is anchored to frequency in training data, which arguably systematically underweights the novel and the rare.

In this sense, LLM reasoning seems to be fundamentally inductive, as the architecture is built on statistical generalisation from observed examples.

The underlying concern seems plausible - AI does appear to struggle with discontinuities and weak signals. However, it is not exactly clear what the authors understand by inductive reasoning. It typically means inductive generalisation - reasoning from some observations to wider generalisations.

The paper explains: “Because AI systems are built upon only existing knowledge, they struggle to identify forward-looking perspectives, unknown/low probability signals and potential disruptions, all of which are necessary for strategic foresight.”

AI in Strategic Foresight AI in Strategic Foresight: Reshaping Anticipatory Governance examines how artificial intelligence is moving from experimentation to practical application in the field of strategic foresight. Developed...

🧵 @oecd-ocde.bsky.social and @weforum.org have published an important paper on #AI in strategic #foresight, surveying 167 foresight experts from 55 countries: www.oecd.org/en/publicati.... One particularly interesting finding is that AI was reported to have limited capacity for inductive reasoning.

what about Mira Murati?

if he wasn’t so incompetent

that’s kinda passive-aggressive, no?😁

How to kickstart the UK economy. With Tim Leunig | The Economics Show Planning and tax reform are at the top of this ex government adviser’s list

Brexit's "great success": £850 million loss a week. That's a number to put on a bus, and a bargain of the century according to @timleunig.bsky.social: pay £350 million - get £850 million. shows.acast.com/the-economic...

so mom of the newly minted Nobel economics laureate Philippe Aghion founded the French fashion house Chloé (and, apparently, even coined the term prêt-à-porter)😳 on.ft.com/487ZfwT

Why Ukraine is winning the war Russia has failed to achieve its core aim — the destruction of the Ukrainian nation

"In 2025, the weakest link in Ukraine’s defences still lies in the minds of its western friends."
www.ft.com/content/2a4d...

thanks to the UN Special Rapporteur Marcelo Vázquez-Bermúdez for including my book (www.cambridge.org/9781316514061) in the 2025 UN report on general principles of law: documents.un.org/doc/undoc/ge.... to clarify - book published by @universitypress.cambridge.org, not @ox.ac.uk as stated there;)

good job, @unesco.org!😉 #aiethics

dx.doi.org/10.1016/j.jr...

or young enough not knowing?

day in honour of @jowolff.bsky.social @blavatnikschool.bsky.social

didn’t know that was a thing🤷🏼‍♂️

just learned that the CEO of Microsoft AI studied philosophy, not tech (dropped out though)

the professor (Adam Smith) and the infidel (David Hume) at the National Portrait Gallery in London.
(with a nod to Dennis C. Rasmussen’s ‘The Infidel and the Professor’ press.princeton.edu/books/hardco...)