undrendo

I'm just a natural philosopher for fun

#OnePunchMan #BusterKeaton #physics #Bayesian #RobertMusil #ItaloCalvino #satori #introvert #misophonia #rstats #UruseiYatsura #anime #manga #linux #Emacs #KDE #dejavufonts #generalrelativity #scihub

Personal website: https://magnaspesmeretrix.org

Last Notes

2026-07-29 07:02:58 UTC

- reply

Thank you for a new brilliant post! LLMs can be very useful in teaching, if they're used as "possible crackpots". I teach BSc physics and statistics; a frequent kind of exercise my students get is to analyse an LLM's output to a physics/statistics question or problem, and determine where it's correct and where it's wrong, explaining *why* those parts are correct or wrong. This puts the students with their back against the wall: they understand that using another LLM to do the exercise would just be running in circles.
This kind of exercise is useful exactly for the reasons you discuss in your post. Some LLM's errors are coarse and evident, but others are very subtle. Examining (first together with the teacher) and looking for this kind of errors is a way to learn and test *discernment* of subtler concepts. Some students even get a sense of empowerment in spotting subtle LLM errors in a subject they're learning – and that's the goal indeed!
Variations of this kind of approach can be made ad libitum. Give the same prompt to two different LLMs, or several times to one LLM; notice the differences; determine which (if any) is correct.
Similarly with summaries: prompt an LLM to summarize a given section or chapter from the textbook, then the students must find out what's wrong, or which important pieces have been omitted, in the output summary.
These kinds of exercise also make the students aware that LLMs are not fully reliable, even when what they're spitting out sounds technical and "expert".
All this also connects to a major pre-LLM problem, nicely expressed by Feynman: "I don't know what's the matter with people: they don't learn by understanding; they learn by some other way – by rote, or something. Their knowledge is so fragile!" (similar sentiments were expressed by many other earlier physicists, like Maxwell). Our education system has favoured memorization instead of discernment, for decades or centuries. "Do we, in our schools and colleges, foster the spirit of inquiry, of skepticism, of adventurous thinking, of acquiring experience and reflecting on it? Or do we place a premium on docility, giving major recognition to the ability of the student to return verbatim in examinations that which he has been fed?" (R. E. Gibson). The appearance of LLMs is, finally, making this rote-learning inadequate, and discernment more and more necessary.
#LLM #education

2025-05-25 11:11:04 UTC

- reply

There's a great guide here: <https://www.big-8.org/wiki/Getting_Started_with_Usenet>, it worked for me.

2025-05-25 08:06:52 UTC

Through the pages of John Baez <https://math.ucr.edu/home/baez> (see also <https://johncarlosbaez.wordpress.com>) I've seen that the sci.physics usenet newsgroup, especially sci.physics.research, always had quite interesting physics discussions; see for instance <https://math.ucr.edu/home/baez/physics>.
Finally I managed to join, but it looks like I'm late to the party. Lots of spam and very insulting posts on sci.physics, and sci.physics.research (luckily a moderated group) seems to have very little activity – unless I'm misjudging the levels of activity of some years ago.
So I'd like to ask @npub1nf4…nqe4 (when/if he has time): what happened to these newsgroups? I'd be curious to hear the story. And which other similar newsgroups or venues are available today? Probably I'm not the only one with these wonderings. Cheers!
#usenet #johncarlosbaez

2025-04-15 16:39:27 UTC

- reply

Well, there are also positive sides. Edwin T. Jaynes <https://bayes.wustl.edu/etj/etj.html> had notorious difficulties in getting his Bayesian-theory papers published, because the "establishment" was largely anti-Bayesian at the time. Luckily he could get published at conference proceedings though.
...Which is a reminder that, in Galilean science, peer-review should be left to the reader – that's why it's called "peer". Otherwise we're back to pre-Galilean times with their Index Librorum Prohibitorum (which included important works by Kepler and Kant). Reviews and commentaries are welcome, but "rejection" (by two or three unknown people, which can scarcely be a representative sample of the whole scientific community) is anti-scientific. Luckily the internet has changed this situation.

2025-04-15 12:40:33 UTC

- reply

There are several steps that look strange:
- eqn (13): the notation "d/dτ" stands for a covariant derivative "u⋅∇", but it isn't clear which connection they're using for the "∇", if it's the same described in the previous lines. But then how is the proper time τ modified, given that they modify the metric to include an electromagnetic potential?
- eqn (24): they define the norm ρ of the charge 4-current to be equal to "-A⋅u". Something feels wrong about the degrees of freedom here. In principle there can be other forces, which means that "A" and "u" can be independent; and we know that "ρ" is independent as well – you can put any charge in any electromagnetic field. This equation instead makes ρ dependent. But maybe I'm miscounting the degrees of freedom.
I think they should have proceeded in a less nonchalant way in their steps and explanations, though, given the potential novelty of this result.
Maybe @npub1nf4…nqe4 has some comments on this paper as well.

2025-04-15 06:11:18 UTC

- reply

Interesting article and approach. From what I see there's no *general* relativity there yet – they are modifying the (Minkowski) metric of flat spacetime. Looking forward to seeing how a general pseudoRiemannian metric and therefore gravity gets integrated into this picture.

2025-01-06 19:27:23 UTC

- reply

According to Weinberg <https://www.harpercollins.com/products/to-explain-the-world-steven-weinberg>, Ptolemy used just one epicycle+referent(+equant) for all planets, except Mercury which needed more. I don't remember if the book discusses Copernicus.

2024-12-14 21:57:21 UTC

- reply

Enclosing relevant pages from Gingerich's book. It'd be cool to redo the calculations Gingerich made.
https://s3.c.im/media_attachments/files/113/653/397/778/986/514/original/5ec19213909443b8.jpg

2024-12-14 19:24:27 UTC

- reply

I'd love to see references where I can verify either claim!

2024-10-08 20:37:37 UTC

- reply

Mine is a firm – and of course personal – idea. I don't see any divisions of subjects either, and in principle I don't care about Nobel prize and similar fluff.
But I'm noticing an increasing problem (a problem from my point of view), which I could express this way: there may be fewer and fewer people who want to hear about ideas from statistical mechanics that underlie machine learning. Because, why bother about statistical mechanics? Just put the data in a machine learning algorithm, and let it make the prediction.
I'm noticing this in my university department and others. As a physicist/mathematician, my deepest pleasure comes from trying to understand a phenomenon, in a combination of physical-geometric visualization and maths; hopefully in a way that leads to new physical-geometric ideas and even new maths. But in many new physics projects there is opposition to this: more and more people push to just put measurement data in machine-learning algorithms, train them, make them predict the quantities of interest. And they may well succeed. But what about the physical understanding of the phenomenon?
Maybe the idea of the Nobel Prize committee was to emphasize the importance of physics to computer science. But I think the result will be the opposite, leading to an increase in the problem above.
Maybe this is just a passing phase. Maybe not. And of course there's no "right" or "wrong". Who says how "physics" should be done? Still it makes me deeply sad.

2024-10-08 19:32:43 UTC

- reply

Agree. Physics ideas have entered in so many different fields. Then one can give a Nobel prize in physics for basically anything. I'd be happy if the Nobel Foundation instituted a Nobel prize in computer science instead.