I knew that some people developed "relationships" with LLMs. I've seen some examples in the press. I could imagine how bad it was. Seeing multiple stories unfolding live on a subreddit is, however, a whole other level of depressing: old.reddit.com/r/MyBoyfrien...

I've revived @mlsimpsons.bsky.social on BlueSky (and hopefully soon on Mastodon). It's currently not scheduled to post automatically, but the pool has some new memes compared to the Twitter account. Feel free to send suggestions in DM! bsky.app/profile/mlsi...

Oh no, the RCE-as-a-service platform is used for malware injection! Who could have known?!

📢 Fully funded PhD - 🌍 Dense Detection of Rare Events in Remote Sensing using Generative Models Leverage generative models, unsupervised segmentation and explainability techniques to map disasters w/ @javi-castillo.bsky.social and Flora Weissgerber Apply ⤵️ recrutement.cnes.fr/fr/annonce/4...

Saviez-vous qu'il existe des règles tacites sur lesquelles s'appuient la numérotation des rues de Paris ? ​🤔 Par exemple, lorsqu'une rue est plus ou moins parallèle aux berges de la Seine, la numérotation croît de l'amont vers l'aval de ce cours d'eau. En savoir plus 👉 www.ign.fr/mag/numerota...

TIL: fitting a monotonic function to an arbitrary curve is called isotonic regression scikit-learn.org/stable/modul...

Sérieusement pour dire « je choisis de faire de la slow science » il faut déjà être dans une position luxueuse (e.g. CNRS) où l’on peut se poser la question et pas à courir partout pour gérer l’enseignement et faire sa recherche dans les interstices

torrentfreak.com/nvidia-conta... corporate ethics is an oxymore

Hi @cvprconference.bsky.social, Could you please for the next year add the checkbox at reviewer registration like "I promise to deliver reviews on time, do NOT spam me everyday like I am stupid and untrustworthy and not adult." Best, Dmytro.

Scholar Inbox is recommending me papers that I'm reviewing. Good job, I guess? 😅

Anthromorphisation des LLM et déshumanisation des personnes atteintes de troubles mentaux. Pas bravo Epsiloon, ni aux auteurs de ce « papier » publié dans un journal prédateur.

Using proper vocabulary is science and teaching 101.

Why bother reviewing then, just reject the paper. 🤨

Why would even write this?

TerraBytes proceedings available in PMLR: proceedings.mlr.press/v292/

Working on image translation? Domain adaptation? Check out FlowEO, newly accepted to #WACV2026. 📝 hal.science/hal-05388176 We introduce a data-dependent flow matching framework to perform image-level unsupervised domain adaptation of remote sensing imagery. Great work from @lebellig.bsky.social!

Signez la pétition demandant la révocation de l'autorisation d'exportation étonnamment accordée à Christie's amis-et-correspondants-centre-international-blaise-pascal.fr/sauvegarde-p...

Apply here: www.ign.fr/nous-rejoind...

🚀 3e succès commercial pour #Ariane6 ! Le lanceur européen a mis en orbite le satellite #Sentinel1D du programme #Copernicus 🌍 Communiqué : cnes.fr/communiques/... #ESA #CNES #Arianespace #ArianeGroup @esa.int @fr.esa.int #VA265

See, this is why the public hates AI. We let these kind of people act like this, because we are spoiled little brats who want bigger and bigger datasets. In the end, this is quite the spitting image of a lot of ML research: "that's just data, dude".

Want to work on generative models and Earth Observation? 🌍 I'm looking for: 🧑‍💻 an intern on generative models for change detection 🧑‍🔬 a PhD student on neurosymbolic generative models for geospatial data Both starting beginning of 2026. Details are below, feel free to email me!

Academics in Assyria in the 7th c BC complain that admin is preventing them from doing research and teaching

The weird thing is that even in academia we fall for this. Every time I have tried an LLM for a real task, it failed, sometimes catastrophically. And I still have colleagues telling me that "I'm not using right" or "not on the right problems". Well, I'm using it for MY problems! And it sucks!

Une nouvelle expo temporaire au meilleur musée de Paris. 🎓

What the hell is that? arxiv.org/pdf/2510.24036 Is this some kind of bibliometric optimization strategy? Students pushing fake plagiarized papers to bootstrap their Google Scholar account? I don't understand what the game is here.

Félicitations au nouveau docteur Maxime Mérizette qui a soutenu avec brio ce matin sa thèse sur la segmentation sémantique de petits objets dans des nuages de points 3D. Un vaste travail mêlant données synthétiques, pseudo-annotations de modèles profonds. Bravo ! 👏🏻

📄 New paper led by @leleogere.bsky.social on detecting notational errors in digital music scores. 🎼 To be presented at TENOR 2025! ⤵️ hal.science/hal-05294807

I'm done with reviewing for AAAI. This year was the cherry on top: 1. Accepted to review 1 paper while in vacation 2. Assigned 4 papers instead of 1 3. Guilt-tripped when I pushed back 4. Assigned an emergency review with 24h notice *on a Sunday* I get it 29000 papers is a lot, but I'm a volunteer!

Breaking news from #AI4EO symposium: cows confirmed as extraterrestrial beings. 👽 #CowsFromSpace More information to come.

I wrote about the tragic death of Adam Raine and the venal negligence of "AI Safety." www.argmin.net/p/the-banal-...

No.

Désappointé par cet article de Next Inpact qui parle de guerres de chapelle dans le monde de « l'IA » et qui mentionne... un unique chercheur, et encore, en passant. On a encore du boulot sur la planche en terme de médiation scientifique. next.ink/196011/lia-s...

I still can't believe the existence of this nerdy heavy metal song about a remote sensing satellite launched by @esa.int Everything about it is so damn perfect. We should have something like that for every single publicly funded project. www.youtube.com/watch?v=wPxO...

I've said before but """agents""" are basically RCE and phishing-as-a-service. I cannot fathom why anyone, moreso anyone with some security awareness, would give an LLM privileged access to their computer.

And now the last keynote after the poster session from Erin Trochim from UAF! Panel session afterwards and best paper awards following that!

After the first block of spotlights we now move to the second keynote by Razvan Cosac from @esa.int Ground Segment!

A #computervision researcher at #ICML2025

Starting with a keynote from Sujit Roy from @thenasaearth.bsky.social IMPACT!

It's almost time! TerraBytes starts tomorrow at @icmlconf.bsky.social 2025 in Vancouver #ICML2025 📍 08:30 West Meeting Room 111-112 🗞️ Papers are now accessible on the workshop website below the programme: terrabytes-workshop.github.io

I'm at #ICML2025 this week. If you wanna chat about ML, geospatial data and Earth observation, or just grab a coffee, hit me up! And come to the TerraBytes workshop on Saturday, we have an amazing lineup on ML for Earth observation! 🌎 terrabytes-workshop.github.io

Pour les collègues francophones, vous saviez que le FID était tout cassé ? Moi non plus. Pourtant, si on s'y prend bien, on peut calculer le FID avec moins de 1000 images. J'en parlerai au GRETSI fin août : hal.science/hal-05142942 👀

𝐓𝐞𝐫𝐫𝐚𝐁𝐲𝐭𝐞𝐬 𝐊𝐞𝐲𝐧𝐨𝐭𝐞 𝐀𝐧𝐧𝐨𝐮𝐧𝐜𝐞𝐦𝐞𝐧𝐭: Razvan Cosac will deliver the second keynote at TerraBytes 2025! Stay tuned for more updates and join us at @icmlconf.bsky.social terrabytes-workshop.github.io

Browsing old mails, I found one from 2015 during my internship at TUM working on facial expression recognition. I had completely forgotten that I was considering using synthetic data at that time! We dabbled with MakeHuman for a while but gave up. Funny I'm still into that 10 years later.

𝐓𝐞𝐫𝐫𝐚𝐁𝐲𝐭𝐞𝐬 𝐊𝐞𝐲𝐧𝐨𝐭𝐞 𝐀𝐧𝐧𝐨𝐮𝐧𝐜𝐞𝐦𝐞𝐧𝐭: Dr Sujit Roy from NASA-IMPACT will deliver the first keynote at TerraBytes 2025! Stay tuned for more updates and join us at @icmlconf.bsky.social 2025

📢 FLAIR-HUB dataset A new large-scale, multimodal dataset for land cover and crop type mapping 🤗 Dataset: huggingface.co/datasets/IGN... 📄 Preprint: arxiv.org/abs/2506.07080 🤗 Pretrained models: huggingface.co/collections/... 💻 Code: github.com/IGNF/FLAIR-HUB 🌐 Project : arxiv.org/abs/2506.07080

📣 Author notifications for TerraBytes are OUT! We've received more than 50 submissions from all around the world. Thanks to our amazing reviewers who have worked on very short notice to help us build an amazing schedule. 🫂 See you all in Vancouver!

➡️ Deadline extension: you can now submit your papers to TerraBytes@ICML 2025 up until May 23. This is the final extension. 😀 Don't pass up this opportunity to present your work at one of the leading machine learning conferences and come represent the geospatial ML community! 🌎

🚨 Workshop alert! 🌍💿 TerraBytes: Towards global datasets and models for Earth Observation Co-located with ICML 2025 (Vancouver, Canada) 🔗 terrabytes-workshop.github.io 📝 We accept short 4-pages and full 8-pages paper on remote sensing, Earth Observation and machine learning. 📆 Deadline: May 16th