Sign in

Gro-Tsen

@gro-tsen.bsky.social
2.6K followers 108 following 18.5K posts

Eclectic mathematician, irritable geek, boring homosexual, procrastinator extraordinaire, overworked dilettante and full-time human being. I post in fr&en.

PostsRepliesMedia
Gro-Tsen @gro-tsen.bsky.social · 2h
Je crois que pour peupler ma base de données PostGIS je les avais récupéré là: www.data.gouv.fr/datasets/adm... (mais comme d'hab c'est un labyrinthe et sur mon téléphone je n'arrive pas à savoir si tu as un .shp dans le tas)
data.gouv.fr
data.gouv.fr
Plateforme des données publiques françaises : télécharger, partager et réutiliser les données de l'État et des collectivités
100 0
Gro-Tsen @gro-tsen.bsky.social · 3h
The terrifying thought is that in 2050 all the people in this montage will be completely typical, average, normal people:
131 0
Gro-Tsen @gro-tsen.bsky.social · 3h
… de légitimité: Torvalds est accepté comme dictateur parce qu'il a fondé le projet, quelqu'un qui serait juste nommé comme dictateur par le fondateur aurait moins de légitimité, donc plus grand risque de fork. Et si Torvalds mourrait subitement demain, ce serait sacrément la merde.
320 0
Gro-Tsen @gro-tsen.bsky.social · 3h
Andrew Morton a 10 ans de plus que Torvalds, et il est toujours très actif, que je sache. Et je ne vois pas pourquoi il faudrait 10 ans pour organiser une transition, ni pourquoi ce serait difficile de trouver qqn de compétent et plus jeune. En revanche, c'est vrai qu'il y a un problème …
100 0
Gro-Tsen @gro-tsen.bsky.social · 17h
I asked a question on MathOverflow about whether we can find a description of the topos of simplicial sets, or of some other topos with the same internal propositional logic, that reflects the description of Medevedev's logic using “finite problems”: mathoverflow.net/q/516110/17064
mathoverflow.net
A topos for Medvedev's finite problems?
I learned from Emil Jeřábek on this site that the internal propositional logic of the topos of simplicial sets is Medvedev logic. The correspondence, however, somewhat obscures Medvedev's original
020 0
Gro-Tsen @gro-tsen.bsky.social · 19h
Grosse promo sur les points d'exclamation!!!!
020 0
Reposted by Gro-Tsen
Lays Farra @laysfarra.bsky.social · 19h
Est-ce que ça se trouve des églises chrétiennes qui chantent les psaumes en hébreu ? À la rigueur j'imagine du côté du judaïsme messianique ("Juifs pour Jésus")
112 0
Gro-Tsen @gro-tsen.bsky.social · 20h
Exilé au Maroc et à la tête d'un cabinet de conseil notamment pour les dirigeants africains: www.gala.fr/l_actu/news_... ❧ www.cnews.fr/france/2026-... ❧ lessentieldeleco.fr/7195-que-dev...
100 0
Gro-Tsen @gro-tsen.bsky.social · 20h
Please feel free to reuse as you will: 🔽
000 0
Gro-Tsen @gro-tsen.bsky.social · 20h
What are the Kleisli and Elienberg-Moore categories on the God monad? 🤔
220 0
Gro-Tsen @gro-tsen.bsky.social · 20h
… I think it's hardly disputable that, at this stage, AI is, on aggregate, causing far more harm than good to mankind, and getting worse, and this is clearly not the fault of the AI's alignment itself but the people driving it, so the previous reasoning does not apply. •8/8
120 0
Gro-Tsen @gro-tsen.bsky.social · 20h
So ultimately, we conclude ℙ(doom) ≪ ¼. But of course all this is conditioned on a superintelligent AI so as to make the “quasi-omnipotent entity” argument sensible. We might have reasons to be much more fearful of doom at human hands (and human stupidity / greed) in the intermediate stages: … •7/8
130 0
Gro-Tsen @gro-tsen.bsky.social · 20h
… which even if this is done imperfectly because “doom” isn't well-specified, should still decrease its probability w.r.t. a completely random alignment which would be just as likely to want something that kills us as something that makes us thrive. •6/8
130 0
Gro-Tsen @gro-tsen.bsky.social · 20h
… if only because it inherited our values through the cascade of intermediate AIs used to produce it, and ultimately through the texts we all wrote, which are pretty clear on the fact that “doom is bad, avoid such outcomes”; … •5/8
130 0
Gro-Tsen @gro-tsen.bsky.social · 20h
… clearly we want doom to be a very undesirable outcome, so as long as we play a role in shaping AI's alignment, it should be repulsive point; and even if a very advanced AI is shaped by a somewhat advanced AI, it should have internalized the desire that “doom is bad”, … •4/8
140 0
Gro-Tsen @gro-tsen.bsky.social · 20h
… if we're going to go by sci-fi authors' ideas. So that places ℙ(doom | “random god”) at maybe ~¼. But of course AI shouldn't be a random god: even if we're going an imperfect job at aligning with our own desires, it should still be much better aligned than random: … •3/8
130 0
Gro-Tsen @gro-tsen.bsky.social · 20h
… (help mankind flourish) or indeed “whimper” (the entity chooses to remain inactive, or maybe self-destruct since there's no reason it should have an instinct of self-preservation), or even weirder eschatological stuff like “merge humans' consciousness into a giant planetary consciousness”, … •2/8
120 0
Gro-Tsen @gro-tsen.bsky.social · 20h
An argument (what “level”?) why ℙ(doom | superintelligent AI) should be small-ish: ⤵️ First, if we bring into existence a quasi-omnipotent entity with ❊randomly❊ aligned desires, there is no reason to believe that “doom” should be higher on its preference list for mankind's fate than “thrive” … •1/8
151 0
Reposted by Gro-Tsen
ill-boo-mi @yallchemist.bsky.social · 23h
who let these people online
An ai slop and entirely incorrect diagram of how generations have different slang to say the same thing
5477 0
Gro-Tsen @gro-tsen.bsky.social · 20h
Everything is stupid.
370 0
Gro-Tsen @gro-tsen.bsky.social · 22h
Et c'est beau en toutes les saisons:
020 0
Gro-Tsen @gro-tsen.bsky.social · 22h
Couleurs d'automne à l'arboretum de Versailles Chèvreloup.
Des arbres dont le foliage est plus ou moins orangeUn cyprès chauve aux couleurs d'automne au bord d'un plan d'eau
151 0
Gro-Tsen @gro-tsen.bsky.social · 22h
Wait, there's both ‘Acer pseudoplatanus’ and ‘Acer platanoides’? Couldn't botanists come up with a better name than “maple tree that looks like a plane tree”? And to top it off: ‘Platanus × acerifolia’, “the plane tree whose leaves look like maple leaves”. 🙄 #ContextClub
020 0
Reposted by Gro-Tsen
Philosophy Now @philosophynow.bsky.social · 16/10/2026
Behold this issue's @existentialcomics.com by Corey Mohler: a very short history of #metaphysics in the Western thought. Explore in our new issue 176: philosophynow.org
715635 0
Reposted by Gro-Tsen
Suave Morbida @maitre-poulard.bsky.social · 16/10/2026
Trop courte carrière de l'emoji coeur poilu! (Google qui a mal interprété les hachures héraldiques, le système de lignes et de points qui permet d’indiquer les couleurs dans une représentation symbolique en noir et blanc)
Heraldic
metal or

U+1 F49B
YELLOVW HEART in
the Unicode reference

Unicode 6.0 used black and white references, therefore it represented yellow and other colors with the hatching pattern of heraldic or. When developing the yellow heart emoji (). Google misinterpreted the dots in Unicode's reference images ()and displayed this emoji as a pink "hairy heart starting from version 4.4 KitkKat. This error was fixed in version 4.4.1 a few months later. l8|9]

Google's emoji, mistakenly
representing the
yellow heart as
hairy
4406 0
Gro-Tsen @gro-tsen.bsky.social · 16/10/2026
Oui, merci! Comme en fait plusieurs personnes m'ont envoyé des versions différentes produites par des modèles locaux, j'en ai plutôt un peu trop, mais je vais essayer de faire une comparaison générale plus tard.
100 0
Gro-Tsen @gro-tsen.bsky.social · 16/10/2026
Sans lire super en détails, les deux versions ont l'air correctes (et essentiellement isomorphes — à des tout petits détails près — à la version que j'ai écrite et illustrée graphiquement).
020 0
Gro-Tsen @gro-tsen.bsky.social · 16/10/2026
La version web de GitHub Copilot fait chier avec les images (il y a plein de limites: sur le nombre d'images attachées, sur le format, sur la taille, etc.).
100 0
Gro-Tsen @gro-tsen.bsky.social · 15/10/2026
… This solution supports a stronger conclusion: if F and G are the intersection with βℝ∖ℝ of cl{sin=0} and {sinᵝ=0} respectively, then: ‣ F has empty interior, ‣ but G contains an open set (in βℝ∖ℝ) containing F. It takes a little bit more work, but I think it's worth it!
020 0
Gro-Tsen @gro-tsen.bsky.social · 15/10/2026
I know everyone else is bored of this exercise by now, but since all 3 AI's had found a solution very similar to mine, I wondered if I could find a substantially different solution, and here's what I came up with (no AI involved here): gist.github.com/Gro-Tsen/bf2... …
Page 1 of a different solution to the same exercise (PDF at link in post)Page 2 of a different solution to the same exercise (PDF at link in post)
110 0
Gro-Tsen @gro-tsen.bsky.social · 15/10/2026
C'est une phrase qui a vraiment été écrite par un would-be infiltrateur, ou c'est une caricature de ce qu'ils sont susceptibles d'écrire?
310 0
Reposted by Gro-Tsen
John C. Baez @johncarlosbaez.bsky.social · 15/10/2026
I can't get any work done today. My chair is nervous, the desk is in a panic... it seems like the whole building is freaking out.
Warnig sign:

This door is alarmed

Photo by John Baez
0271 0
Gro-Tsen @gro-tsen.bsky.social · 15/10/2026
… J'ai vraiment du mal à voir quel serait la raison pour laquelle ce serait stupide de vouloir enregistrer systématiquement quel modèle d'IA, et quel prompt d'entrée, a servi à générer tel commit, par exemple au cas où on découvrirait après coup que tel modèle tend à faire telle catégorie d'erreurs.
220 0
Gro-Tsen @gro-tsen.bsky.social · 15/10/2026
… Du coup, la comparaison avec l'utilisation de compilateurs, ce ne serait pas «c'est interdit de s'en servir, vous devez écrire votre code assembleur à la main» mais «vous devez écrire dans un commentaire quel compilateur a servi à générer ce code assembleur» et ça ne me semblerait pas débile. …
110 0
Gro-Tsen @gro-tsen.bsky.social · 15/10/2026
Je ne comprends pas du tout: le corps de l'article parle de (la fin de) l'obligation de signaler l'usage d'IA dans un commit, mais le message de Torvalds a l'air de se moquer de l'interdiction tout court d'utiliser l'IA dans un commit, ce qui a l'air d'être une question complètement différente. …
100 0
Gro-Tsen @gro-tsen.bsky.social · 15/10/2026
Mainly because LLMs (in their current incarnation) don't seem to have much of a clue of what we find “interesting”, or even easy/hard. But I'm not defending the position that there's no way LLMs can be used productively in mathematical research: I'm saying that currently, they're a net negative.
100 0
Gro-Tsen @gro-tsen.bsky.social · 15/10/2026
I think I need to post this once again:
252 0
Gro-Tsen @gro-tsen.bsky.social · 15/10/2026
“It has 20% chance of leading to humanity's extinction. It's already super dangerous and could help bioterrorists and cyberattacks. It's on the verge of becoming an indestructible computer virus. And it gets used to stalk, harass and attack online. But why don't you like us? 🤔”
152 0
Gro-Tsen @gro-tsen.bsky.social · 15/10/2026
These tech billionaire AI bros certainly know how to sell their product by making it sound desirable. 😬
150 0
Gro-Tsen @gro-tsen.bsky.social · 14/10/2026
Zut, ça me rappelle tout d'un coup une conversation sur la différence de sens des mots “draco” et “serpens” en latin, j'aurais dit que c'était sur Bluesky, et peut-être avec toi, mais je ne retrouve plus.
000 0
Gro-Tsen @gro-tsen.bsky.social · 14/10/2026
Oh great! Then let me tell you I love it. But I didn't realize it was autobiographical. 😅
020 0
Gro-Tsen @gro-tsen.bsky.social · 14/10/2026
I feel like I've just seen the real-life version if this joke. 😁
Comic following the “no space in brain” meme template suggesting homotopy theory takes all the place in one person's brain, who goes on to rant that the sphere spectrum is a more natural object than the integers
150 0
Gro-Tsen @gro-tsen.bsky.social · 14/10/2026
Je ne pense pas, mais n'hésite pas à essayer de déchiffrer si tu penses que si!
000 0
Gro-Tsen @gro-tsen.bsky.social · 14/10/2026
Street art, rue Gérard, Paris
Black and white street art on a square-tiled surface
391 0
Gro-Tsen @gro-tsen.bsky.social · 14/10/2026
Commentaires de l'acteur (Bruce Campbell) qui a joué ce rôle: www.hollywoodreporter.com/movies/movie...
hollywoodreporter.com
‘Escape From L.A.’ at 25: Bruce Campbell Revisits Creepy Cameo While Marveling How Film’s Satire Became Reality
The John Carpenter sequel bombed at the box office, but has some great moments which helped it become a cult classic.
050 0
Gro-Tsen @gro-tsen.bsky.social · 14/10/2026
Further comments about this graph: 🔽
100 0
Gro-Tsen @gro-tsen.bsky.social · 14/10/2026
… Even though I was able to prove that {sinᵝ = 0} in βℝ is larger than the closure in βℝ of {sin = 0} (and contains an open neighborhood of it in the corona), I couldn't picture it. But part (C) in the above drawing helps me get some sense of “why” this holds.
000 0
Gro-Tsen @gro-tsen.bsky.social · 14/10/2026
Of course the corona βℝ∖ℝ is a monstrous object as it embodies “all possible behaviors at infinity of all possible continuous real-valued functions”. We can only picture continuous images of it; yet they “suffice”, and this is what this graph tries to illustrate. …
100 0
Gro-Tsen @gro-tsen.bsky.social · 14/10/2026
The whole point of the original exercise 🔽 was for me to try to gain an intuitive grasp of βℝ (or β[0,+∞[ but no matter). Knowing the definition of a mathematical object (or even proving things about it), and having an intuitive grasp, are different things.
100 0
Gro-Tsen @gro-tsen.bsky.social · 14/10/2026
(The image was generated by this Python+matplotlib script gist.github.com/Gro-Tsen/af2... — itself initially generated by ChatGPT from very precise instructions of mine about what to draw where, and then further edited by me.)
gist.github.com
stone_cech_plot.py
GitHub Gist: instantly share code, notes, and snippets.
100 0