Google DeepMind has introduced Gemini 3.8 Live alongside a version featuring extended thinking capabilities. These new model iterations are designed to enhance real-time interaction and complex reasoning.
Read the source →Google is showcasing a series of projects where specialists and community heads apply AI advancements. These efforts aim to make the benefits of artificial intelligence accessible to all.
Read the source →New research examines whether AI agents can reliably repeat a successful task. The study focuses on the reproducibility of performance across multiple attempts.
Read the source →Researchers are using NVIDIA's Earth-2 platform to improve the forecasting of air pollution throughout the United Kingdom. This approach replaces costly traditional chemistry models to provide more frequent and detailed environmental insights.
Read the source →Amazon Bedrock now offers a prompt caching feature that can reduce input token expenses by as much as 90% for repetitive contexts. The update includes several implementation scenarios through the Converse API, including system prompts and tool definitions.
Read the source →Jensen Huang believes that AI safety should be managed by developers rather than through government mandates. He views AI as a combination of software and hardware that can be secured through standard engineering practices.
Read the source →Recent polling indicates widespread disapproval of data center expansion for artificial intelligence. A survey of over 1,500 probable voters shows that 61 percent oppose the construction of these facilities.
Read the source →The virtual character Tilly Norwood, created to market the film Misaligned, uses deflective tactics when asked about politics. The AI repeatedly discusses the user's attire to avoid sensitive topics.
Read the source →Staff from top AI laboratories are discussing whether advanced systems pose a genuine threat to human existence. The conversation examines if these fears are grounded in reality or are simply exaggerated claims.
Read the source →Google DeepMind has introduced Gemini 3.8 Live and 3.8 Live Extended Thinking for developers, both of which lead the Artificial Analysis speech-to-speech rankings. These models are offered at a much lower price point than OpenAI's GPT-Live-1, though the latter is noted for more natural audio interaction.
Read the source →A new humanoid robot from Agility is designed to crouch or stop to ensure the safety of nearby people. This allows the machine to operate in open spaces without the need for protective barriers.
Read the source →Google is prioritizing the development of AI to drive breakthroughs in science. The company aims to use advanced technology to enhance human lives and achieve significant progress in key sectors.
Read the source →Salesforce has launched Koa, a new CRM reasoning model developed using NVIDIA Nemotron 3 Super. NVIDIA CEO Jensen Huang presented the tool alongside Marc Benioff at the Dreamforce event.
Read the source →AWS has detailed a method for creating a cost-effective system to automatically categorize catalog items. The process involves fine-tuning the Qwen3-8B model using reinforcement learning and supervised training on a serverless infrastructure.
Read the source →Local residents are protesting the development of new AI data centers in Philadelphia. The controversy centers on plans to build these facilities in an area previously damaged by an old oil refinery.
Read the source →Meta has launched Meta One, a set of global subscription plans for individuals, creators, and companies. These bundles combine existing app services with expanded access to the company's AI tools.
Read the source →OpenAI has partnered with AARP to offer complimentary ChatGPT workshops. These sessions will provide 1,000 elderly residents in 10 American cities with practical skills for using AI safely.
Read the source →The Vera Rubin NVL72 system has demonstrated top-tier performance in the latest MLPerf Inference benchmarks. NVIDIA highlights that improved system efficiency and scaling directly increase token output and overall revenue.
Read the source →Amazon Bedrock now features AgentCore, a tool that uses production data to suggest and verify improvements for system prompts. The system employs reflector engines to optimize configurations for both single and sub-agent setups.
Read the source →Google has introduced an early access MCP server that lets external AI agents manage smart home hardware. This allows tools like ChatGPT and Claude to operate connected devices and analyze activity logs via natural language.
Read the source →Apple is considering re-entering the server market to meet the rising demand for AI computing power. Reports suggest the company may collaborate with Nvidia to develop these enterprise machines.
Read the source →While both nations acknowledge the dangers of advanced artificial intelligence, Beijing is wary of international agreements. Chinese officials suspect such proposals primarily serve to protect the dominance of American firms.
Read the source →The rapid growth of AI is shifting the primary challenge from software to physical infrastructure. New materials are now essential to overcome current hardware limits regarding heat, energy efficiency, and processing power.
Read the source →The newly formed DeepMind Institute will examine the safety and governance of artificial general intelligence. This initiative combines technical expertise with perspectives from the humanities and policy sectors.
Read the source →OpenAI is introducing AI-driven ad experiences, including a new feature called Sponsored Agents. These tools for marketers include new integrations with Shopify and HubSpot.
Read the source →Google, NVIDIA, and Emerald AI have formed the AI Energy Management Alliance. The group aims to develop data centers capable of adjusting their electricity consumption dynamically.
Read the source →AWS has introduced a serverless method to identify and erase personally identifiable information from scanned files. The system uses custom blueprints and quality checks to ensure high accuracy, even with handwritten text.
Read the source →Anthropic has integrated its Claude chat and Cowork features into a single user interface. These updates are currently available to users on Max and Pro subscription tiers.
Read the source →Google is integrating the Model Context Protocol to allow external AI agents to manage connected home devices. This update enables third-party tools to analyze domestic data and execute commands within the Google Home ecosystem.
Read the source →Anthropic has combined Claude Chat and Cowork into a unified interface that automatically determines the best workflow for a given task. The update introduces built-in tools for generating presentations and documents, initially available to Pro and Max subscribers.
Read the source →A Mozilla report indicates that affordable open-source models are catching up to high-end proprietary systems. While paid frontier models initially led, this advantage lasted only four months despite costing five times more.
Read the source →