Understand the signal before you apply.
AI releases, research, funding shifts and application guidance—checked against cited sources and connected to current global opportunities.
352 published insights · page 17 of 20
MIT η-Learning Generates Plausible Extreme Events Without Examples
MIT's Extreme Event Aware η-learning generates statistically plausible rare-event maps even when the training subset contains few or no extreme examples.
IBM Reveals 2nm Dual-Architecture Processor for Z and LinuxONE
IBM previewed a 2nm processor whose cores are designed to execute IBM and Arm instruction sets while adding AI acceleration for future Z and LinuxONE systems.
AWS Adds Managed Ray Workflows to SageMaker HyperPod for Training and Inference
AWS has added a managed Ray experience to SageMaker HyperPod on EKS, including Studio-based cluster management, Ray observability, resilient training, tiered checkpointing and Ray Serve infe...
Google ADK Adds Native Evaluation for Live and Voice AI Agents
Google's Agent Development Kit now supports native live-agent evaluation with simulated audio users, multi-turn scoring, tool-trajectory checks and CI/CD automation for voice-agent regressio...
SLAC AI Compression Preserves Fine Scientific Data at 10–100× Smaller Size
SLAC researchers report a neural compression method that can shrink scientific datasets by roughly 10–100× while preserving fine-scale features and enabling selective decompression.
Apple IVT Cuts Visual Chain-of-Thought Latency by More Than 5×
Apple researchers introduce Internalized Visual Thinking, a post-training method that learns future visual representations during training while avoiding image generation at inference.
UK and Ukraine Sign Defence AI Partnership With Pilot-First Deployment
The UK and Ukraine signed a non-binding AI partnership covering co-developed models, secure data and compute, assurance, research and pilot-first defence deployment.
Anthropic CHIVE Tests LLM Explanations with Counterfactual Prompt Edits
Anthropic Fellows Program researchers introduced CHIVE, an agentic pipeline that discovers unexpected model behaviors and tests proposed explanations with measured counterfactual prompt edit...
Google DeepMind Expands Games Research From Atari to EVE Online
Google DeepMind is building on 15 years of game-based AI research by partnering with game developers to prototype new gameplay experiences and research environments.
Google DeepMind Expands AI-and-Games Research from Atari to EVE Online
Google DeepMind is extending its long-running use of games as AI research environments by working with game developers on new gameplay prototypes, including work connected to EVE Online.
NVIDIA Maps Where Security Should Live in an AI Agent Stack
NVIDIA argues that agent security needs authoritative runtime and infrastructure controls below modifiable models and harnesses, with least privilege, isolation, scoped credentials and audit...
Google Research Uses Mobility Patterns to Improve AI Understanding of Places
Google Research introduced a mobility-informed framework that adds temporal activity patterns to place representations, improving model predictions of attributes such as opening hours, price...
NVIDIA Draws a Security Boundary Below the AI Agent Harness
NVIDIA's AI safety and security teams argue that agent harnesses should guide behavior while runtimes and infrastructure enforce identity, policy, credentials, isolation and auditability.
DeepSeek V4 Flash Vision Exp Brings Multimodal Agents to Its API
DeepSeek released an experimental multimodal V4 Flash variant with image input, agent support, three API interfaces and a reusable Files API.
Anthropic CHIVE Tests Whether Interpretability Explains LLM Behavior
Anthropic's CHIVE pipeline uses counterfactual prompt experiments to test explanations of real model behaviors and finds no uplift from three activation-reading tools.
Anthropic Research Finds Fine-Tuned Lie Detectors Generalize Poorly
Anthropic-affiliated researchers found that fine-tuned self-report lie detectors improved in-distribution accuracy but transferred poorly to unseen deception types.
ITU-T Standardizes Low-Latency, Energy-Efficient Network Framework for AI
ITU-T approved Recommendation Y.3335, a framework for low-latency, energy-efficient networks across operators that Japanese telecom companies link to future physical-AI and all-photonics inf...
Meta Details Muse Spark 1.2 Multimodal Reasoning, Robotics and WildArtifactBench
Meta AI Research published new Muse Spark 1.2 demonstrations and evaluation details spanning visual coding, robotics, audio-video understanding and a preview of WildArtifactBench.
Turn intelligence into applications.
Create a free alert for the topics, roles or countries that matter. We email only new verified matches.
More ways to save
Discover deals, coupons and free courses on our sister site.
