1854 stories tagged AI · page 7 of 78
A new arXiv paper argues that AI agents should be evaluated like biological systems through systematic observation and perturbation, not just performance outcomes. This shift could lead to more reliable and understandable AI behavior.
Replit has introduced Free Mode, powered by GPT-5.6 Luna, allowing anyone to create software without worrying about token costs. This move democratizes software development by removing financial barriers.
A systematic review of 260+ studies on ArXiv details how large language models (LLMs) are applied in mental health — from social media analysis for early depression detection to clinical conversational agents and therapy support tools — while highlighting key ethical challenges around privacy, consent, and reliability.
Flock Safety's new AI tool, OS Investigate, helps police analyze surveillance footage and license plate data to solve crimes. WIRED obtained and reviewed the tool's source code, revealing advanced facial recognition and vehicle tracking capabilities that raise significant privacy concerns.
Do-over is a tool that creates snapshots of files before AI agents like Claude Code execute commands, allowing you to undo accidental deletions. It protects untracked files and those outside your git repository, filling a gap that git alone can't cover.
Asana used OpenAI's Codex to automate a testing system that would have taken five years to complete manually. The project cost about $12,000 and was finished in just two weeks.
Archron is a new tool that allows AI agents to update customer relationship management (CRM) systems like Salesforce and HubSpot. It creates an immutable audit log of all changes, ensuring accountability and security.
Stripe is reportedly acquiring OpenRouter, an AI gateway startup, for over $7 billion. This deal highlights the growing importance of AI infrastructure in financial services. OpenRouter's technology allows developers to easily integrate multiple AI models into their applications, similar to how Stripe simplifies payment processing.
OpenAI has introduced a new initiative to enhance democratic oversight of AI in national security, providing government institutions with tools, training, and expertise to ensure responsible and transparent use of AI in sensitive areas.
Nvidia is investing $1.5 billion in a SoftBank subsidiary that builds data centers, guaranteeing its chips will power an upcoming OpenAI facility. The deal locks in Nvidia's hardware for one of the most anticipated AI data center projects.
OpenAI has introduced a dedicated ChatGPT mode for teenagers, combining existing youth safeguards with new safety features like stricter content filters and more robust moderation. The launch comes amid mounting public scrutiny over how AI tools affect younger users.
OpenAI launched ChatGPT for Teens, a version of its AI chatbot designed for younger users. It includes stronger safety features and parental controls to help teens learn and use AI responsibly.
OGX (Open GenAI Stack) is an open-source AI application server and Python library that implements the APIs of major AI labs including OpenAI, Anthropic, and Google. It allows developers to build agentic AI applications using a single API surface, then deploy with any combination of inference engine, vector database, and safety backend without changing application code.
Meta released Glimmer, an open-weight AI model that anyone can download and run on their own hardware. This contrasts with Meta's more powerful Muse Spark, which remains locked behind its APIs. Mark Zuckerberg argues AI should be accessible to everyone, not just a few labs.
Google is now allowing users to turn off visible watermarks on AI-generated images. Invisible watermarks for identification will remain in place.
Anthropic will use invisible watermarks based on Google DeepMind's SynthID-Text to comply with EU AI transparency rules. These watermarks embed detectable patterns in Claude's text using word probabilities, enabling identification of AI-generated content without affecting readability.
A new benchmark study shows that large language models can calibrate their confidence to match the strength of medical evidence, a key step toward more reliable AI-assisted clinical decision-making.
Writer has introduced a new AI model based on Z.ai's open-source GLM-5.2, designed to reduce deployment costs. The model includes an upgraded 'harness' to manage token expenses efficiently.
A new ArXiv study reveals that large language models give confident wrong answers due to stable miscalibration, where errors persist under small input perturbations. The research introduces an audit score and internal sensitivity probe to diagnose this systematic overconfidence.
NVIDIA released MegaTrain, a tool that lets researchers train AI models with up to 100 billion parameters on a single H100 GPU, potentially reducing training time and cost by up to 50%.
Researchers developed a new method to induce judging rubrics from data, reducing the tendency of AI judges to over-credit fluent but unsuccessful agent performances. This improves reliability over hand-written rubrics like G-Eval or fine-tuned weights.
MegaTrain, an open-source tool from developer Wast3, enables training 100-billion-parameter AI models on a single consumer GPU, dramatically lowering the hardware barrier for large-scale AI development.
MBZUAI, Cerebras, and Inception have released Jais 2, a family of open large language models optimized for Arabic. The family includes a 70-billion-parameter model, the largest open Arabic-centric LLM trained from scratch, and an 8-billion-parameter variant. Both models use a custom Arabic-centric vocabulary for efficient training and inference.
Andrew Ng has published a detailed map of the most important skills for AI engineers, covering machine learning, data engineering, software engineering, and soft skills. The resource provides a clear roadmap for professionals and newcomers entering the field.