Jonas Meeuws
Ludovic CourtèsIn case you missed it: https://www-cs-faculty.stanford.edu/~knuth/papers/claude-cycles.pdf
Anyone knows of a good farming 101 tutorial?
Andrew Helwer@civodul lmfao irreversible psychic damage
bartosz 🚲🌳🐍🦀
Siddhesh Poyarekar@civodul this looks like the exact same use case as the OpenAI press release about "a new discovery in physics". Which, AFAICT seems similar to me, to the protein folding problem that LLMs proved effective at tackling.
Basically all of these problems appear to be that of searching the language space for a general solution pattern that can be validated against a bunch of specific results. LLMs seem quite effective at converging to valid solutions compared to simple brute forcing.
@siddhesh_p @civodul right, LLMs are known to be good in pattern matching: https://arxiv.org/abs/2601.11432
The unreasonable effectiveness of pattern matching
We report on an astonishing ability of large language models (LLMs) to make sense of "Jabberwocky" language in which most or all content words have been randomly replaced by nonsense strings, e.g., translating "He dwushed a ghanc zawk" to "He dragged a spare chair". This result addresses ongoing controversies regarding how to best think of what LLMs are doing: are they a language mimic, a database, a blurry version of the Web? The ability of LLMs to recover meaning from structural patterns speaks to the unreasonable effectiveness of pattern-matching. Pattern-matching is not an alternative to "real" intelligence, but rather a key ingredient.
G. Clavier@siddhesh_p @civodul The important point every enthousiast is missing is that Claude provided a "code solution" that worked for "a restrain number of use case" (my words, not Knuth’s quotes). Knuth wrote the actual formal solution himself based on piece of code he rewrote. This very different from saying "Claude solved a mathematical problem" it also took it 31 iterations before providing a valid solution, with elusive computing ressources.
Woke Leftist Trash@civodul is it that surprising? Any function f can be simulated by a neural net within some epsilon p, given enough resources to run the network.
It’s not much different than alpha go, right? You keep tweaking the model, exploiting circuits and other behaviours, intercepting and rewriting prompts, re-prompt, align, etc until you get it closer to the epsilon for the given problem domain.
It’s not like the algorithm has a theory of mind and, knowing it’s audience, write an elegant proof. It’s still doing what LLMs do: regurgitating statistically likely next tokens. It just happens to have regurgitated something Knute expected on this one execution.
gkrnours
Pjotr Prins@civodul Heh. Unless you choose to use just use a rake and hoe and sell your produce by hand, farming is even more tainted and locked in by propietary solutions these days. Even manure is a propietary industry now. We may be on a sliding slope, but be careful what you ask for :).
Simon Tournier@civodul Yeah, as @pjotrprins said, farming is also tainted these days.
Therefore, I would root a “good farming 101 tutorial” by some “Amish Technology”. 🙃
« Reinforcing Values and Building Community »
🤔
Short article published in 2007 (IEEE Technology and Society Magazine)
https://web.archive.org/web/20170830063308id_/https://classes.soe.ucsc.edu/cmpe080e/Spring08/Amish.pdf
dave@zimoun @pjotrprins Just yesterday I glanced at a documentary on Amish life:
https://www.arte.tv/fr/videos/111822-000-A/les-amish-la-vie-au-passe/
Getting inspiration. :-)
Les amish, la vie au passé - Regarder le documentaire complet | ARTE
Originaires de Suisse et d'Allemagne, les amish ont émigré en Amérique du Nord au XVIIIe siècle. Ils sont aujourd'hui connus pour leur mode de vie singulier, marqué par le conservatisme religieux et par le rejet de la modernité.
Cayetano Santos
yak@civodul very surprising that the program worked for general m. I am used to people saying that these problems are easy for LLMs to solve for fixed or bounded m, but that they don't generalize well. But the methodology here was very different from those and arguably closer to how I would approach the problem.
Thomas@civodul One of the greatest minds in computer science, who has seen every AI wave yet, expressed an informed opinion on the current wave. I find it abhorrent for Lispers in particular to dismiss him out of hand. Maybe Knuth is full of it this time, but that's really the bet you're going to make?!?!