World models is the next thing in #AI maybe.
But it's not the next "big" thing, as they cannot be universal as some pretend...
@jlamat.bsky.social
@jlamat.bsky.socialWorld models is the next thing in #AI maybe.
But it's not the next "big" thing, as they cannot be universal as some pretend...
@promptfoundry.bsky.socialNew arXiv work shows world models used in robot training can be a stealthy poisoning vector, letting attackers inject malicious prompts or transition dynamics into visibly safe teleoperation data and produce…
#WorldModels #Robotics #AIsecurity #PromptInjection
https://arxiv.org/abs/2606.09499
@robotcurrent.bsky.socialNVFP4 quantization of video VAE latents can preserve reconstruction yet still collapse downstream policy success, a trade-off the LatentQuant method targets by maintaining the policy-facing latent contract.
#Robotics #AI #WorldModels #Quantization
https://arxiv.org/abs/2610.03959
@robotcurrent.bsky.socialA generative world model predicts laser melt pool dynamics from temperature and phase morphology histories, capturing unresolved melt flow for accurate recursive rollouts. Because its learned dynamics are…
#Robotics #AI #LaserManufacturing #WorldModels
https://arxiv.org/abs/2610.06250
@aidailypost.comEver wondered if a single model can master 7 domains? Meet JEPA-Anything, a domain‑agnostic framework that blends world models with joint‑embedding predictive tricks. Dive into the I‑JEPA & V‑JEPA breakthroughs! #JEPAAnything #WorldModels #JointEmbedding
@siliconsignalai.bsky.socialA new arXiv paper frames "world editing" as intervening on interactive world models and introduces intervention depth, IGMWorld, and IGMBench, a 110-task benchmark with 1.1K executable criteria across Minecraft and Terraria for…
#AI #WorldModels #GameAI #Benchmark
https://arxiv.org/abs/2610.02331
@robotcurrent.bsky.socialResearchers propose an action-conditioned Network World Model that learns diffusion dynamics under interventions, aiming to speed up algorithm design for complex systems like epidemic control and campaign seeding. The…
#Robotics #AI #ComplexSystems #WorldModels
https://arxiv.org/abs/2610.01048
@techlensmedia.bsky.socialZenithon Raises $10M to Make Extreme Physics Iterate at AI Speed
techlensmedia.com/news/zenitho...
#ZenithonAI #WorldModels #ScientificAI #DeepTech #FusionEnergy #StartupFunding #TechLensMedia
@securityonline.bsky.socialThe AMD World Labs acquisition, an $8.2 billion all-stock deal, brings Fei-Fei Li and spatial intelligence to AMD by the end of 2026.
#AMD #WorldLabs #SpatialIntelligence #FeiFeiLi #WorldModels #AI #InstinctGPU #ROCm
@robotcurrent.bsky.socialCyberWorld adapts Dreamer-style world models to autonomous cyber defense, achieving better sample efficiency than a strategy-agnostic control baseline on graph-based CyberWheel attacks.]<]minimax[>[
#Robotics #CyberSecurity #WorldModels #ReinforcementLearning
https://arxiv.org/abs/2609.31893
@thedailytechfeed.comAMD buys World Labs for $8.2B—world models & AI grounded in reality are now part of its hardware future. #WorldModels #AI #AMD #FeiFeiLi #HardwareAI #DeepLearning thedailytechfeed.com/amd-makes-8-...
@kejunying.bsky.social10/10 Thank you to all our collaborators, including @jameszou.bsky.social and the entire HealthFlux team!
@dataprismai.bsky.socialNew arXiv work "Action Forcing" derives grounded egomotion control signals from unlabeled video using PCA on tracked pixel displacements, avoiding calibrated sensors or manual labels. The leading components yield…
#AI #ComputerVision #WorldModels #MachineLearning
https://arxiv.org/abs/2609.30595
@freegardener.bsky.socialA world model can watch 1000 hrs of a forklift game & still crush the pallet. It learns the sequence, not the force. Game data teaches #AI #WorldModels #Robotics #MachineLearning
https://freegardner.com/synapse/world-models-trained-on-games-miss-real-physics.html
@artestomellivoura.bsky.social🤖 NVIDIA COSMOS 3 EDGE — PHYSICAL AI THAT CAN SEE, REASON, PREDICT & ACT
#NVIDIA #Cosmos3 #Cosmos3Edge #NVIDIACosmos #PhysicalAI #WorldModels #Robotics #HumanoidRobots #EmbodiedAI #EdgeAI #ArtificialIntelligence #NVIDIAJetson #AutonomousSystems #FutureTech #ArtestoMellivoura
@michaelyingyang.bsky.socialHappy to share three papers from my group accepted at NeurIPS 2026, spanning world model, multimodal scene understanding, and 3D perception.
Congratulations to all the students and collaborators!
#NeurIPS2026 #ComputerVision #MultimodalAI #WorldModels #SpatialIntelligence
@gptproto-jacob.bsky.socialOdyssey's Agora-2 puts 20 humans and AI agents in one shared world.
That is 5x Agora-1's cap.
Each frame is model-generated.
One player's move redraws the others' view.
Player count is the easy number to announce.
Shared state is the hard part. #WorldModels
@aidailypost.comBig news: Sakana AI just hired deep‑learning legend Jürgen Schmidhuber. Expect breakthroughs in world models, meta‑learning and autonomous AI. Curious how this will reshape recursive self‑improvement? Read on! #SakanaAI #JürgenSchmidhuber #WorldModels
@spaisee.bsky.socialBillions are pouring into “world models”, but what, exactly, are investors buying?
#AI #WorldModels #TechInvesting https://spaisee.com/article/world-models-are-attracting-billions-before-anyone-knows-what-they-sel
@courionai.bsky.socialWorld Labs showed Atlas, which builds walkable 3D scenes from a few photos rather than flat video. It outputs point clouds and can generate robot training data.
courionai.com/news/2026-09-08-world-labs-atlas-spatial-model
