Simon Friederich

AGI is not inevitable, and if you are American, you may consider getting active to stop its development.

Great piece by @garrisonlovely.bsky.social.

In contrast, crying "AI hype" makes people complacent and only plays into the hands of the industry.

If the answers were in the training data, it would indeed be far less impressive. I am assuming that they aren't.

Apparently, the computing power used was quite extraordinary, though. Still, the effort necessary to achieve this kind of thing will also come down.

Alexander Wei on X: "1/N I’m excited to share that our latest @OpenAI experimental reasoning LLM has achieved a longstanding grand challenge in AI: gold medal-level performance on the world’s most prestigious math competition—the International Math Olympiad (IMO). https://t.co/SG3k6EknaC" / X 1/N I’m excited to share that our latest @OpenAI experimental reasoning LLM has achieved a longstanding grand challenge in AI: gold medal-level performance on the world’s most prestigious math competition—the International Math Olympiad (IMO). https://t.co/SG3k6EknaC

Again, so much for "AI hype":

An OpenAI experimental reasoning LLM has performed at gold medal level at the International Math Olympiad (IMO):

x.com/alexwei_/sta...

(To be clear, I personally find this quite depressing.)

Very nice teaser.

Thanks, Ingrid (if I may), amazing reply! I should have addressed you directly. I will read more by you and perhaps react if I feel I have anything to contribute with potential substance.

My worries about advocacy apply in particular to policy suggestions I myself have made in the past.

Yes, those are definitely very important issues.

It's certainly an idea that should be studied, but there are reasons that economists are profoundly sceptical. Their objections may not be the last word, but I would be a little more comfortable if research on and advocacy of limitarianism were more clearly separated.

Inside the Secret Meeting Where Mathematicians Struggled to Outsmart AI The world's leading mathematicians were stunned by how adept artificial intelligence is at doing their jobs

So much for "hype":

'Ono says. “I don’t want to add to the hysteria, but in some ways these large language models are already outperforming most of our best graduate students in the world.”'

www.scientificamerican.com/article/insi...

I find the monoculture aspect almost uniformly bad. I came over because of Musk and the changes he made there, but I find it much more boring here than Twitter was, and it doesn't have anything remotely resembling the rich (and often quite nuanced) AI content that X has. Can you understand?

Gefeliciteerd. Goed voor CDA en voor Nederland. Een premier die eerder geduldig en zonder ideologie academische literatuur over technische aspecten van de energietransitie aan het brede publiek heeft uitgelegd, zou een uitstekende keuze zijn.

Nodig is: minder inkomensverschillen, meer ontmoeting in de publieke ruimte en meer onbevangenheid, aldus @tkupp.bsky.social @rug.nl #werkaandewinkel

Oh, how embarrassing I didn't realize.

Very useful! Could you add me?

YouTube video by TED The AI Revolution Is Underhyped | Eric Schmidt | TED

The AI revolution is underhyped -- perhaps not everywhere, but certainly on Bluesky.

www.youtube.com/watch?v=id4Y...

New paper, in Ergo: To the extent that we are sometimes cluelessness about the more indirect consequences of our actions, this speaks against a longtermist orientation, not in favour: journals.publishing.umich.edu/ergo/article...

(The thesis may seem banal, but Hilary Greaves argued the opposite.)

Yes.

And the correct response to worries about the (power concentrating aspects of) AI is to push for regulations that limit increases in capabilities, not to refrain from use.

It seems to me that as educators we will have to do two things

1 Do evaluations of students without AI (paper and pencil tests)

2 Teach students how to use AI to be more productive

The notion that we can totally shun AI in higher ed when it will be so important is misguided.

To some extent it is already being fixed, see here for instance: futureagi.com/blogs/unders...

More generally, the more AI systems are deployed agentically, the more they will run against adverse consequences of mistakes, including hallucinations. The time scale is unclear, but it will happen.

Hallucinations are going down. Medium-term there are strong incentives to fix this entirely. Don't count on this remaining a weakness of generative AI indefinitely.

Are there any other academics here on Bluesky who are also very worried about AI, in particular AI-driven power concentration, who do *not* think that the technology is overhyped?

For academics that are skeptical about AI's ability to provide perceptive & useful results, a fast way to test your assumptions is to give one of your papers as a PDF to o3 (or Gemini 2.5 but it lacks some tools), with the prompt:

"critique this paper"

I have found very good accuracy & insight...

For Silicon Valley, AI isn’t just about replacing some jobs. It’s about replacing all of them | Ed Newton-Rex AI will do the thinking, robots will do the doing. What place do humans have in this arrangement – and do tech CEOs care? says Ed Newton-Rex, founder of Fairly Trained

Fantastic piece by @ednewtonrex.bsky.social:

www.theguardian.com/commentisfre...

AI companies aim at automating the entire economy, preparing to take away all our bargaining power.

They may well succeed unless we coordinate to impose restrictions on AI capabilities.

Bluesky is a much nicer environment than X, but the AI conversations seem much more superficial here. The very stimulating new contributions on AI as normal technology, gradual disempowerment, and the intelligence curse, to name just a few, have been discussed more lively on X. Very sad.

These are good, Ken. You will not be surprised to hear that I agree almost completely. 8 is just hard for me (lack of competence), and I tend to disagree with 6. Focusing on specific phenomena rather than on recovering why the formalism in its entirety works doesn't seem promising to me.

Unsolicited advice for making progress in quantum foundations:

1) If there were a gentle tweak or simple idea that would make clear sense of quantum theory, beyond the for-all-practical-purposes-level, it would certainly have been found by now.

Gefeliciteerd, Maarten!

Waar is de tijd dat progressieven nog in vooruitgang en verlichting geloofden! Zoals Kamagurka ooit zei: "Vroeger was de toekomst beter". Maar tegenwoordig is 'vooruitgang' bijna een vies woord voor links. Hoog tijd voor een progressieve beweging die haar naam wél weer waarmaakt! Nu in de winkel.

Screenshot of the latest editorial page from the Journal for General Philosophy of Science. The title reads "Editorial" with authors Helmut Pulte and Thomas Reydon listed below. The article was accepted on 14 February 2025. A DOI link (https://doi.org/10.1007/s10838-025-09727-7) is displayed at the top.

In our latest editorial, we express our sincere thanks to Guido Bacciagaluppi for his editorial work since 2021 & bid him a fond farewell as he concludes his tenure. We also warmly welcome @simonfriederich.bsky.social as our new Co-Editor-in-Chief! link.springer.com/article/10.1... #philsci #philsky

YouTube video by The Quietist A Mechanism for Entanglement?

Latest version of my work on entanglement with @kenwharton.bsky.social. We propose that entanglement is a kind of selection artefact. This was a talk at @griffith.edu.au, and the slides are accessible here: bit.ly/4h9dMeo

#philsky #PhilSci #quantum #entanglement

Heel goede inbreng. Genuanceerd en toch heel duidelijk. Dank voor al je werk en inzet.