- Sabine's Newsletter
- Posts
- CERN’s Plans, Copenhagen, Watching AIs Think, and Cosmological Crises
CERN’s Plans, Copenhagen, Watching AIs Think, and Cosmological Crises
This week’s science bits from SWTG
Sabine Hossenfelder
April 02, 2025
|
How Large Language Models Think
Example of how Claude navigates potentially harmful requests. Source: Anthropic.
A super interesting study from Anthropic looked at the reasoning chains of Claude 3.5 Haiku for various tasks. One of the examples is the addition that Claude does by using a sort of heuristic approximation using known results. When asked, however, it will claim that it did it with mental arithmetic similar to that which humans use. Another example is how hallucinations happen. It seems to be a failure of the model to recognize that it does not know the answer to a question. Read the full study here.
This week’s episode of Science News is about cosmology. The branch of physics that deals with the universe is facing a bit of a crisis right now — four crises, actually. Let’s take a look at the biggest problems with our current understanding of the universe and how we got here to begin with. This week’s video also comes with a quiz, which you can take here.
Speaking of quizzes, you can now create and share your own quizzes on QuizWithIt for free! Each quiz has a unique URL, can be embedded into websites or newsletter, and be shared on social media. Happy quizzing!
Opposition to CERNs Mega-Collider Plan Increases
Image: CERN
CERN is expected to put forward detailed plans for their Future Circular Collider (FCC) in the next few weeks, and journalists finally found some physicists willing to speak out against this insanity. I mean, other than me. In an interview with Nature, particle physicist Ruben Saakyan says that “People inside the community do say [the FCC] is really the current [Director General’s] vision which has been pushed forward.” According to the same article: “Some researchers told Nature they felt pressured to back the FCC to help present a unified front to the outside world.”
This does not remotely surprise me, seeing as I have been accused of being “anti-science” for pointing out that physics would be better off if the money were invested into more promising research areas like quantum information or new accelerator technologies that could shrink particle colliders back down to a reasonable size.
The FCC, which would likely cost more than $40 billion, wouldn’t be in full operation until the mid 2070s. In my opinion, it would be a huge mistake to tie so much money into a megaproject that is unlikely to benefit anyone besides particle physicists, and by the time it’s built might be obsolete already.
CERN’s annual funding is not sufficient to finance the FCC. They are planning to increase member state contributions, but still need several billion dollars on top of that. It is currently unclear where that money is supposed to come from.
A member of the German ministry for science and education said already last year that Germany isn’t going to pick up the bill, quite possibly because Germany is choking on their own collider project that is 10 years beyond schedule and by now a stunning €3 billion above budget. Last month, 400 French scientists signed a petition against the FCC.
Copenhagen Still Rules
A survey at a physics conference last year with 85 respondents revealed that among the interpretations of Quantum Mechanics, the Copenhagen interpretation is still the most popular, with 28% favoring it, followed by Many Worlds with 11% (though I would argue that “no opinion,” with 33%, is as popular as the Copenhagen interpretation). The survey also found that hybrid models of dark matter (some combination of modified gravity and particle dark matter) have become the most popular candidate. Paper here.
Seeking impartial news? Meet 1440.
Every day, 3.5 million readers turn to 1440 for their factual news. We sift through 100+ sources to bring you a complete summary of politics, global events, business, and culture, all in a brief 5-minute email. Enjoy an impartial news experience.
