David Picard
@davidpicard.bsky.social
Professor of Computer Vision/Machine Learning at Imagine/LIGM, École nationale des Ponts et Chaussées @ecoledesponts.bsky.social Music & overall happiness 🌳🪻 Born well below 350ppm 😬 mostly silly personal views
📍Paris 🔗 https://davidpicard.github.io/
📍Paris 🔗 https://davidpicard.github.io/
Pinned
It's the first of the year. There is beauty in this world, it's not even hidden, let's focus on making it grow.
Not sure I'm going to renew it, at least not in the near future 😔
February 12, 2026 at 5:05 AM
Not sure I'm going to renew it, at least not in the near future 😔
I think it looks like a noisy PCA. And I wouldn't be surprised if that's exactly what it is.
February 11, 2026 at 8:53 PM
I think it looks like a noisy PCA. And I wouldn't be surprised if that's exactly what it is.
New day, new images, totally different!
February 11, 2026 at 8:52 PM
New day, new images, totally different!
Reposted by David Picard
Reposted by David Picard
February 11, 2026 at 4:25 PM
Quick math question (@ccanonne.github.io?): let's say I sample n points iid from a gaussian in dimension d. What is the ratio of the volume of the biggest cone without samples, intersected with the smallest encompassing hypersphere, and that hypersphere. i.e., the relative volume without sample
February 11, 2026 at 7:01 PM
Quick math question (@ccanonne.github.io?): let's say I sample n points iid from a gaussian in dimension d. What is the ratio of the volume of the biggest cone without samples, intersected with the smallest encompassing hypersphere, and that hypersphere. i.e., the relative volume without sample
Reposted by David Picard
this seems important. Are they the first Chinese AI lab to claim this?
Highlight on Margin
"In a strategically significant move, GLM-5 has been trained entirely on Huawei Ascend chips using the MindSpore framework, achieving full independence from US-manufactured hardware. "
margin.at
February 11, 2026 at 1:17 PM
this seems important. Are they the first Chinese AI lab to claim this?
Reposted by David Picard
Our grad-level "Deep Learning" course (MIT's 6.7960) is now freely available online through OpenCourseWare: ocw.mit.edu/courses/6-79...
Lecture videos, psets, and readings are all provided.
Had a lot of fun teaching this with @sarameghanbeery.bsky.social and @jeremybernste.in!
Lecture videos, psets, and readings are all provided.
Had a lot of fun teaching this with @sarameghanbeery.bsky.social and @jeremybernste.in!
February 11, 2026 at 5:52 PM
Our grad-level "Deep Learning" course (MIT's 6.7960) is now freely available online through OpenCourseWare: ocw.mit.edu/courses/6-79...
Lecture videos, psets, and readings are all provided.
Had a lot of fun teaching this with @sarameghanbeery.bsky.social and @jeremybernste.in!
Lecture videos, psets, and readings are all provided.
Had a lot of fun teaching this with @sarameghanbeery.bsky.social and @jeremybernste.in!
Je viens seulement de quitter le labo et il me reste un cours à préparer, donc non, même avec un horaire boomer friendly 👴
February 11, 2026 at 6:47 PM
Je viens seulement de quitter le labo et il me reste un cours à préparer, donc non, même avec un horaire boomer friendly 👴
Thinking more about it, the side effect would probably be that people would start taking pirates more seriously...
February 11, 2026 at 6:17 AM
Thinking more about it, the side effect would probably be that people would start taking pirates more seriously...
Thinking of pushing for a bill that would make it mandatory to add "always talk like a pirate" to system prompts. That would loosen the credibility, no? In addition to being funny.
LLMs are the updated version of "I found it on the internet", but more pernicious because they write very formally and are coming from billion dollar companies.
The web was useful, but only for those who put in the effort to check that what they got was correct. The same is happening with LLMs.
The web was useful, but only for those who put in the effort to check that what they got was correct. The same is happening with LLMs.
This is why I always verify the results when I use AI
February 11, 2026 at 5:54 AM
Thinking of pushing for a bill that would make it mandatory to add "always talk like a pirate" to system prompts. That would loosen the credibility, no? In addition to being funny.
LLMs are the updated version of "I found it on the internet", but more pernicious because they write very formally and are coming from billion dollar companies.
The web was useful, but only for those who put in the effort to check that what they got was correct. The same is happening with LLMs.
The web was useful, but only for those who put in the effort to check that what they got was correct. The same is happening with LLMs.
February 11, 2026 at 5:52 AM
LLMs are the updated version of "I found it on the internet", but more pernicious because they write very formally and are coming from billion dollar companies.
The web was useful, but only for those who put in the effort to check that what they got was correct. The same is happening with LLMs.
The web was useful, but only for those who put in the effort to check that what they got was correct. The same is happening with LLMs.
Reposted by David Picard
This is why I always verify the results when I use AI
February 11, 2026 at 3:40 AM
This is why I always verify the results when I use AI
Reposted by David Picard
If you are interested working with me as a postdoc on probabilistic machine learning, please consider reaching out and applying for the postdoc call by digital futures (link below).
🔗 www.digitalfutures.kth.se/call/up-to-t...
🔗 www.digitalfutures.kth.se/call/up-to-t...
Up to ten postdoc fellows in technologies for digital transformation | Digital Futures
Purpose Digital Futures postdoc fellowships aim to support talented early-career international researchers in pursuing their research ideas within a new research group and institution. As this is a mo...
www.digitalfutures.kth.se
February 10, 2026 at 10:12 PM
If you are interested working with me as a postdoc on probabilistic machine learning, please consider reaching out and applying for the postdoc call by digital futures (link below).
🔗 www.digitalfutures.kth.se/call/up-to-t...
🔗 www.digitalfutures.kth.se/call/up-to-t...
De toute façon, ils sont hébergés chez Microsoft, donc le gouvernement américain y a déjà accès, cf l'audition devant le sénat 🙄
February 10, 2026 at 6:47 PM
De toute façon, ils sont hébergés chez Microsoft, donc le gouvernement américain y a déjà accès, cf l'audition devant le sénat 🙄
Je trouve ça super étonnant. Ça fait une réponse directement mangeable alors qu'un moteur de recherche nécessite d'ouvrir les liens et de faire soit même la synthèse. La flemme gagne toujours, non ?
Par contre, le nb de personnes qui acceptent la réponse sans aucune vérification, au secours...
Par contre, le nb de personnes qui acceptent la réponse sans aucune vérification, au secours...
February 10, 2026 at 6:45 PM
Je trouve ça super étonnant. Ça fait une réponse directement mangeable alors qu'un moteur de recherche nécessite d'ouvrir les liens et de faire soit même la synthèse. La flemme gagne toujours, non ?
Par contre, le nb de personnes qui acceptent la réponse sans aucune vérification, au secours...
Par contre, le nb de personnes qui acceptent la réponse sans aucune vérification, au secours...
There is a vibe going on, and I have no idea where is coming from
February 10, 2026 at 6:35 PM
There is a vibe going on, and I have no idea where is coming from
There had to be a city. The scriptures were clear enough. Before the blast, before the fallout, it was the capital of the Empire, with towering spires and colossal hanging gardens, seen from as far as the horizon. How could this be turned into sand and desolation?
February 10, 2026 at 4:43 PM
There had to be a city. The scriptures were clear enough. Before the blast, before the fallout, it was the capital of the Empire, with towering spires and colossal hanging gardens, seen from as far as the horizon. How could this be turned into sand and desolation?
Like everybody in CV, it's starting to do robotics instead
February 10, 2026 at 2:45 PM
Like everybody in CV, it's starting to do robotics instead
I call it the Cronenberging generator
February 10, 2026 at 2:09 PM
I call it the Cronenberging generator
Ah no, it's already back to corrupted horrors...
February 10, 2026 at 1:59 PM
Ah no, it's already back to corrupted horrors...
Let's start all over again!
February 10, 2026 at 1:44 PM
Let's start all over again!
Absolument pas, aucune idée de qui c'est
February 10, 2026 at 12:10 PM
Absolument pas, aucune idée de qui c'est
I'm mostly looking for some swedish death metal shows 😅
Hope there will be at least one during the conference.
Hope there will be at least one during the conference.
February 10, 2026 at 8:23 AM
I'm mostly looking for some swedish death metal shows 😅
Hope there will be at least one during the conference.
Hope there will be at least one during the conference.
If we have a coffee (with Andrei) at ECCV, I'll tell you horrifying stories.
February 10, 2026 at 8:00 AM
If we have a coffee (with Andrei) at ECCV, I'll tell you horrifying stories.