home.social

#worldmodels — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #worldmodels, aggregated by home.social.

fetched live
  1. A world model can be well-calibrated one step at a time and confidently wrong a hundred steps out.

    New tutorial — The Full Loop:

    • PILCO solved real cart-pole in 17.5s by planning through the posterior
    • Ensemble disagreement misses compounding rollout error (Biased Dreams, RLC 2026)
    • Conformal bounds beat asking the model how sure it is
    • Abstention is a first-class action

    artifocial.com/blog/probabilis

    #WorldModels #AI

  2. A world model can be well-calibrated one step at a time and confidently wrong a hundred steps out.

    New tutorial — The Full Loop:

    • PILCO solved real cart-pole in 17.5s by planning through the posterior
    • Ensemble disagreement misses compounding rollout error (Biased Dreams, RLC 2026)
    • Conformal bounds beat asking the model how sure it is
    • Abstention is a first-class action

    artifocial.com/blog/probabilis

    #WorldModels #AI

  3. Un bon article d'introduction à JEPA (Joint-Embedding Predictive Architecture), famille d'architectures IA appartenant au domaine des World Models pronée par Yann LeCun #IA #WorldModels loud-technology.com/blog/jepa-

  4. Un bon article d'introduction à JEPA (Joint-Embedding Predictive Architecture), famille d'architectures IA appartenant au domaine des World Models pronée par Yann LeCun #IA #WorldModels loud-technology.com/blog/jepa-

  5. 🚀✨ Wow, Ars Technica bravely tackles the bold ambition of "simulating everything"—or maybe just a tiny bit, if we're being honest. 🤔 In a groundbreaking exposé, they reveal the *scandalous* limits of world models, which we totally didn't see coming. Who knew reality was so pesky? 🙄
    arstechnica.com/ai/2026/07/sim #ArsTechnica #SimulationLimits #WorldModels #TechNews #ScandalousReality #HackerNews #ngated

  6. 🚀✨ Wow, Ars Technica bravely tackles the bold ambition of "simulating everything"—or maybe just a tiny bit, if we're being honest. 🤔 In a groundbreaking exposé, they reveal the *scandalous* limits of world models, which we totally didn't see coming. Who knew reality was so pesky? 🙄
    arstechnica.com/ai/2026/07/sim #ArsTechnica #SimulationLimits #WorldModels #TechNews #ScandalousReality #HackerNews #ngated

  7. World Models are enabling AI to simulate and predict physical environments, unlocking new capabilities in robotics, simulation, and planning. Our latest blog — Beyond Text AI: How World Models Are Reshaping Artificial Intelligence — breaks down what world models are, why they matter, and where they’ll have the biggest impact. Read the full article: wix.to/AS3bi8v

    #AI
    #Research
    #Innovation
    #WorldModels
    #MachineLearning

  8. World Models are enabling AI to simulate and predict physical environments, unlocking new capabilities in robotics, simulation, and planning. Our latest blog — Beyond Text AI: How World Models Are Reshaping Artificial Intelligence — breaks down what world models are, why they matter, and where they’ll have the biggest impact. Read the full article: wix.to/AS3bi8v

    #AI
    #Research
    #Innovation
    #WorldModels
    #MachineLearning

  9. winbuzzer.com/2026/06/01/nvidi

    NVIDIA has launched Cosmos 3 as a physical-AI model that combines scene reasoning, multimodal generation and action output, tying the release to a new OpenMDW licensing framework.

    #AI #NVIDIA #PhysicalAI #AIModels #MultimodalAI #WorldModels #Robotics

  10. winbuzzer.com/2026/06/01/nvidi

    NVIDIA has launched Cosmos 3 as a physical-AI model that combines scene reasoning, multimodal generation and action output, tying the release to a new OpenMDW licensing framework.

    #AI #NVIDIA #PhysicalAI #AIModels #MultimodalAI #WorldModels #Robotics

  11. LeCun Launches Broadside Against LLMs, Questions Hinton's Stance

    AI pioneer Yann LeCun questions Large Language Models and promotes world models for AI's future. Learn why he left Meta and started AMI.

    #YannLeCun, #LLM, #AIMachineLearning, #WorldModels, #AIFuture

    newsletter.tf/yann-lecun-criti

  12. AI expert Yann LeCun says current AI models like LLMs are not on the path to true intelligence. He believes 'world models' will be the future of AI within five years.

    #YannLeCun, #LLM, #AIMachineLearning, #WorldModels, #AIFuture
    newsletter.tf/yann-lecun-criti

  13. David @npub1cfhh50298407nqc9pf2ahdn5dcxuxkzhpextg07rrv49tzsyzz5sq7khav@momostr.pink ·