Turing Award winner Yann LeCun, who left Meta last year, has founded AMI Labs in Paris to build world models based on the Joint Embedding Predictive Architecture (JEPA), backed by $1 billion in invest…
A developer trained a Joint-Embedding Predictive Architecture (JEPA) world model, called LeMario, on Super Mario Bros from scratch, finding the model could predict five-step futures and perform reward…
Researchers introduced Physics-Grounded Symbolic Architecture (PGSA), a new AI model that achieves exact linear identifiability across all physical regimes, overcoming the Gaussian limitation of Joint…
Elon Litman published a technical walkthrough of Joint Embedding Predictive Architectures (JEPA), Yann LeCun's self-supervised learning method that trains models to predict representations of masked i…
GeoLambda GmbH released Oikoumene v0.3.1, an autonomous agent civilization simulator that models human history from 70,000 years ago to 2100 using JEPA world models and a Maslow-style needs hierarchy.…
Yann LeCun's AMI Labs raised $1.03 billion at a $3.5 billion valuation to build AI systems that understand the physical world, arguing that current large language models cannot achieve human-level int…
Researchers trained a Joint Embedding Predictive Architecture (JEPA) on unlabeled driving data from the nuPlan dataset and used temporal prediction error as a zero-shot complexity score, finding it ef…
Yann LeCun's AMI Labs raised $1.03B at a $3.5B valuation to build world models using JEPA, signaling a shift from LLMs to physics-aware AI. NVIDIA's Cosmos platform, trained on 20 million hours of rea…
Turing Award winner Yann LeCun told Bloomberg that large language models will become largely obsolete within five years, arguing that scaling text-based AI cannot achieve human-level intelligence. LeC…
Harold Hotelling's 1936 Canonical Correlation Analysis (CCA) is the conceptual foundation for modern Joint Embedding Predictive Architecture (JEPA) models, as JEPA implicitly performs a nonlinear gene…
Researchers at arXiv propose a hybrid pre-training objective for language models that combines a Joint Embedding Predictive Architecture (JEPA) latent-space prediction loss with masked language modell…
DeepSeek V4 made million-token context cheap, Yann LeCun argued for JEPA world models over LLMs, and Lovable demonstrated hourly self-improvement loops for AI agents. The Tokenizer edition #30 also co…