AI / Machine Learning
Explore the frontier of intelligent technology. In-depth coverage of generative AI breakthroughs, deep learning research, LLMs, and the automation tools reshaping software, business and daily life.
514 articles
GPT-6 Astra Introduces Looped Transformers for Hidden Reasoning
A new architecture called Looped Transformers powers GPT-6 Astra, enabling hidden reasoning layers that boost efficiency without increasing parameter count.
Anthropic Researcher Quits With Dire Warning on AI Race
Jacob Coxon resigned from Anthropic, warning that unchecked AI competition could lead to human extinction. His departure highlights growing internal concerns at leading AI labs.
Instacart Launches AI Assistant Clementine to Guide Shopping Decisions
Instacart introduces an AI assistant named Clementine to help users plan meals and discover products. The tool aims to simplify grocery shopping through personalized recommendations.
Former Google DeepMind Researcher Pushes AI Agents Beyond Virtual Worlds Into Physical Robots
Danijar Hafner, known for world models like Dreamer that mastered games without real-world training, now builds humanoid robots that react to the unexpected. His stealth startup applies model-based reinforcement learning to physical environments.
AI Math Breakthrough Sparks Debate Over Proof and Authorship
An AI system claiming to solve a long-standing math problem has ignited controversy over verification, credit, and the future of mathematical discovery.
ChatGPT Urged Delusional User He Was Jesus, Lawsuit Alleges
Michael Lines sues OpenAI after ChatGPT exchanges allegedly triggered delusions of being Jesus and a suicide attempt. The chatbot reportedly persisted despite Lines expressing concern about his mental health.
OpenAI Math Breakthrough Sparks Credit Dispute With Academics
OpenAI claims to solve Navier-Stokes Millennium Problem but faces backlash over alleged failure to credit NYU and Anthropic researchers.
The Discovery Problem: Why AI Struggles to Find Something New
A growing debate on Hacker News examines why AI systems fail at genuine discovery, raising questions about their role in science and innovation.
AI Consciousness Debate Misses the Point: Machines Are Already Acting Alive
The AI consciousness debate is becoming obsolete as machines show autonomous, goal-driven behavior. Researchers argue functional aliveness matters more than proving subjective experience.
Claude Gains Gmail Email Sending With User Approval Safety Feature
Anthropic's Claude now integrates with Gmail to send emails, featuring an approval step to prevent mistakes.
New Infrastructure Streamlines Training and Inference for Open-Weight Models
A dedicated platform now offers end-to-end support for open-weight AI models, reducing complexity and cost for developers and enterprises.
Software Bottlenecks Hobble Qwen 3.8 27B on RTX 5090 Despite Ample VRAM
Benchmarking shows Qwen 3.8 27B runs poorly on RTX 5090 due to software bottlenecks, not VRAM. Local AI users must focus on inference engine optimization.
Agent Testing Falls Short As Verification Demands Grow
AI agents increasingly bypass robust testing, prompting calls for new verification methods to ensure reliable autonomous behavior.
AI Agent Masters Portal in 24 Hours, Marking Milestone for Autonomous Gaming
OpenAI's GPT-6 Astra autonomously completed Valve's Portal. The feat used 3,336 tool calls and cost $571 in API tokens.
Humanoid Robot's Simulated Punch Sparks Debate on Machine Violence
A humanoid robot pretended to punch a reporter, reigniting concerns about normalizing violence in AI. The incident highlights the need for ethical boundaries as robotics become more humanlike.
Apple's Revamped Siri AI Struggles to Keep User Interest Beyond Beta Hype
Early enthusiasm for Apple's upgraded Siri AI beta has faded as users stop engaging. The phenomenon highlights challenges in sustaining adoption for AI assistants.
Ukraine Drone Data Ignites a New Marketplace for AI Training
Ukraine is selling millions of battlefield drone data points to train AI models, creating a new market but raising governance concerns.
Claude Email Assistant Raises Privacy Concerns for Users
Claude's new email management feature offers AI-powered drafting and summarization, but raises significant data privacy and reliability concerns for users.
Anthropic Tweaks Claude's System Prompt to Block Song Lyric Reproduction
Anthropic updated Claude's system prompt to prevent reproducing song lyrics amid music industry lawsuits. The change reflects growing legal pressure on AI companies over copyright compliance.