today's ai news

📰 AI NEWS 1. 🧠 AMD Makes an $8.2B Bet on Spatial AI 📰 WHAT HAPPENED !! AMD agreed to acquire World Labs in an $8.2 billion all-stock deal , bringing Fei-Fei Li’s spatial-intelligence company into AMD. World Labs develops models that can reconstruct and generate interactive 3D environments from visual information. The transaction connects advanced AI-model research with the hardware, software and systems needed for robotics, simulation and physical-AI workloads . After closing, Li is expected to join AMD as executive vice president and chief scientist. World Labs’ research team will continue developing spatial-intelligence models while becoming part of AMD’s broader AI infrastructure strategy. 🧠 WHAT CAN IT DO? World Labs builds spatial-intelligence models that generate, reconstruct and simulate interactive 3D environments from text, images and video. Its technology also supports robotic learning and simulation, giving AI systems a way to work with richer representations of...

📰 AI NEWS 1. 🛡️ NVIDIA Builds a Safety Fence Around AI Agents 📰 WHAT HAPPENED !! NVIDIA launched the NVIDIA Open Agent Safety Platform , an open software platform and reference system designed to control AI agents from testing through deployment. It combines OpenShell , a security-focused runtime, with NVIDIA Sentry , a reference design that watches agent behavior from outside the workload. OpenShell can trace actions and enforce policies, while Sentry uses a separate watchdog layer to detect boundary violations and isolate an agent. NVIDIA says the approach covers software, compute and robotics, and is intended to give developers enforceable controls rather than relying only on model-level instructions or application guardrails. 🧠 WHAT CAN IT DO? The platform can sandbox agent workloads, trace actions, enforce permissions and watch behavior at runtime . Its Sentry reference design adds an external watchdog that can quarantine an agent when it attempts to cross defined boundaries...

📰THIS WEEK'S  AI NEWS 1.  👓 PrismML Puts a Tiny AI Model Inside Smart Glasses 📰 WHAT HAPPENED !! PrismML demonstrated a tiny vision-language model running locally on Qualcomm-powered smart glasses , bringing AI inference closer to the device instead of relying entirely on remote servers. The 2-billion-parameter Bonsai model is optimized for Qualcomm’s Snapdragon AR1 Gen 1 platform and is designed to understand what the wearer is seeing in real time. PrismML’s broader approach focuses on shrinking models while retaining much of their benchmark performance. The demonstration matters for wearable AI because local inference can reduce network dependence, improve responsiveness and potentially offer stronger privacy for cameras and microphones operating continuously around a user. 🧠 WHAT CAN IT DO? The compact Bonsai model can process visual and language information directly on compatible hardware. That creates a path toward wearable assistants that can understand surro...