MIT-IBM Watson AI Lab students enhance AI trustworthiness with innovative probes and techniques, ensuring more reliable and robust models. Their research focuses on understanding and improving the uncertainty of large learning models, crucial for applications like IBM's Granite Guardian family.
NVIDIA CEO Jensen Huang and chief scientist Bill Dally honored for pioneering AI architectures, revolutionizing computing industry. Their work recognized with 2025 Queen Elizabeth Prize for Engineering, highlighting essential role of AI in future progress.
MIT researchers at CSAIL are pioneering a modular approach to coding with large language models, breaking systems into clear "concepts" and "synchronizations" to create more transparent and understandable software. This new method, presented at the Splash Conference, aims to address the challenge of feature fragmentation in modern systems by simplifying how different pieces of functionality int...
New Tsetlin Machine binary classification tech predicts species with high accuracy using unique algorithm. Demo showcases strength and weakness of binary encoding in machine learning.
UNESCO adopts global standards on ethics of neurotechnology, responding to AI advances and consumer neurotech devices. Effort aims to regulate the "wild west" of brain data technologies.
Amazon introduces AgentCore Gateway, a centralized tool server for AI agents to access and invoke specialized Model Context Protocol (MCP) servers. This unified interface streamlines tool management, authentication, and routing, breaking down silos for enterprise teams in managing multiple MCP servers.
'Vibe coding' named Collins dictionary's 2025 word of the year, combines AI with software development. Lexicographers use 24bn-word Collins Corpus to track emerging words like 'biohacking' and 'broligarchy'.
Amazon Search optimized GPU instance utilization by leveraging AWS Batch for SageMaker Training jobs, increasing peak GPU-accelerated instance utilization from 40% to over 80%. The managed solution allowed for advanced prioritization criteria, seamless integration, and increased efficiency in model development at Amazon Search.
MIT researchers have developed a new AI-driven system for search-and-rescue robots to quickly generate accurate 3D maps of complex scenes. The system can process an arbitrary number of images in real-time, making it easier to scale up for real-world applications in disaster scenarios and industrial settings.
Artificial intelligence fuels rise in online harassment against women like Gaatha Sarvaiya, creating fear of distorted deepfake images. Sarvaiya, an Indian law graduate, faces challenges building her public profile due to potential misuse of her posted images.
Amazon sues Perplexity AI for accessing customer accounts and disguising AI as human browsing in shopping feature. Amazon's lawyers demand an end to Perplexity's misconduct, emphasizing the unauthorized access and deceptive practices.
Google search results displayed false information from an AI website about drivers risking a $250 fine for not keeping their headlights on at all times. The New South Wales transport department warns of misinformation about road rules due to the rise of AI, with false claims spreading online.
Global stock markets drop as AI company valuations cool. Bank bosses warn of possible market correction ahead after record highs.
Google plans to launch AI datacenters into space by 2027, using solar-powered satellites to meet increasing demand for AI processing. Engineers aim to deploy about 80 satellites in orbit 400 miles above Earth, taking advantage of falling rocket launch costs.
Amazon Bedrock AgentCore offers two deployment methods: container-based and direct code deployment (for Python), providing flexibility for developers. Direct code deployment simplifies the process by packaging code and dependencies into a zip archive, benefiting those who prefer not to deal with Docker expertise.