Our vision is for AI that uses world models to adapt in new and dynamic environments and efficiently learn new skills. We’re sharing V-JEPA 2, a new world model with state-of-the-art performance in visual understanding and prediction. V-JEPA 2 is a 1.2 billion-parameter model,
MACHINE LEARNING
-
Autonomous AI Drone Beats Human Pilots in World Race
By
–
Saviez vous qu'il existe des compétition de conduite de drones ?
— VISION IA (@vision_ia) 11 juin 2025
Et pour la première fois, un drone autonome piloté par une intelligence artificielle a battu les meilleurs pilotes humains lors d’une course mondiale.
La victoire du MAVLab de l’Université de Technologie de Delft… pic.twitter.com/FnSkXVavpqSaviez vous qu'il existe des compétition de conduite de drones ? Et pour la première fois, un drone autonome piloté par une intelligence artificielle a battu les meilleurs pilotes humains lors d’une course mondiale. La victoire du MAVLab de l’Université de Technologie de Delft
-
Turning Cutting-Edge AI Papers Into Real-World Products
By
–
Since 2023, I’ve helped developers and data scientists turn cutting-edge papers into real-world products.
-
Developer Implements Entire Diffusers Library from Scratch in R
By
–
this absolute maniac implemented the entire diffusers library… from scratch… in R didnt know that was possible (does it run on GPU?). and i dont know if i love it or hate it. but god do i respect it
-
USA vs UK: AI and Biosciences Ecosystem Comparison
By
–
In case you’re wondering, ChatGPT can give a pretty good summary of what is going on in the UK in comparison to the USA in AI and biosciences. ChatGPT: Here’s how the USA and UK compare across AI and biosciences, focusing on company count, startups, and market capitalization:
-
AI-Powered Flood Forecasting for Resilient Communities Worldwide
By
–
Global flood forecasting requires a lot of data that isn't always available. But with AI trained to understand the relationship between rainfall and streamflow, we can make flood predictions around the world which is a crucial step for building more resilient communities
-
AI Self-Improves by Rewriting Its Own Source Code
By
–
Je viens d’analyser un papier où une IA… s’améliore toute seule.
— VISION IA (@vision_ia) 10 juin 2025
Elle code, échoue, se corrige, réécrit son propre code source, et relance le cycle.
Un véritable darwinisme algorithmique.
Extrait : pic.twitter.com/o21y8MrYagJe viens d’analyser un papier où une IA… s’améliore toute seule. Elle code, échoue, se corrige, réécrit son propre code source, et relance le cycle. Un véritable darwinisme algorithmique. Extrait :
-
Nano-Transformers: Simplifying Model Implementation Without Dependencies
By
–
huggingface transformers remains extremely useful for getting started but has bloated beyond recognition and is no longer helpful for learning someone should make nano-transformers. implement all the models from scratch without all the dependencies. would be amazing
-
RL Training Paradigm: 14B Model Matches 32B Baseline Performance
By
–
New training paradigm: instead of just predicting tokens, models reason about each prediction using RL The model thinks through context, considers alternatives, then makes a prediction. 14B model matches 32B baseline, though training costs are significantly higher.
-
Next Generation GPTs: Hybrid Models Beyond Autoregressive Approaches
By
–
it is increasingly likely that the next generation of GPTs will not be vanilla autoregressive models, nor text diffusion, but some third hybrid thing if you believe this, read subham's research. this paper makes one step in that direction (making diffusion work with kv cache)