News/AI Models
Cognizant’s AI Lab reaches 59 U.S. patents with neural network breakthroughs
Cognizant's AI Lab has secured its 59th U.S. patent, marking a significant milestone in the company's artificial intelligence research efforts. The achievement reflects Cognizant's accelerating innovation pace, with two new patents granted in the first half of 2025 alone, plus an additional 23 patents pending approval. What you should know: The latest patents demonstrate Cognizant's focus on solving core AI challenges around neural network optimization and training efficiency.• U.S. Patent No. 12,282,845 covers Multi-objective Coevolution of Deep Neural Network Architectures, designed to improve model performance and resource efficiency across applications from medical imaging to natural language processing.• U.S. Patent No....
read Jul 23, 2025OpenAI and Oracle expand Stargate project with 4.5 GW capacity boost
OpenAI and Oracle are expanding the Stargate AI infrastructure project with an additional 4.5 GW of data center capacity, bringing the total U.S. development to over 5 GW. This massive buildout advances OpenAI's commitment to invest $500 billion in 10 GW of AI infrastructure over four years, while creating over 100,000 jobs across construction, operations, and related industries. What you should know: The expansion represents a significant leap in AI infrastructure development, with real operations already underway. The combined capacity will power over 2 million chips across Stargate facilities, designed specifically for AI workloads that traditional data centers can't handle....
read Jul 22, 2025New method tracks how AI models actually make predictions after scaling
AI researcher Patrick O'Donnell has introduced "landed writes," a new method for understanding how large language models make predictions by tracking how internal components actually influence outputs after normalization scaling. The approach addresses a critical gap in current AI interpretability tools, which measure what model components intend to write rather than what actually affects the final answer after the model's internal scaling processes. The core problem: Most AI interpretability tools completely miss how transformer models internally reshape component contributions through RMSNorm scaling, which can amplify early-layer writes by up to 176× while compressing late-layer contributions. When a neuron writes +0.001...
read Jul 22, 2025Alibaba’s Qwen3 model outperforms rivals while cutting hardware costs by 70%
Alibaba has released Qwen3-235B-A22B-2507-Instruct, an open-source large language model that outperforms rival Chinese AI startup Moonshot's Kimi-2 and Claude Opus 4's non-thinking version on key benchmarks. The model comes with an FP8 version that dramatically reduces compute requirements, allowing enterprises to run powerful AI capabilities on smaller, less expensive hardware while maintaining performance quality. What you should know: The new Qwen3 model delivers substantial improvements across reasoning, coding, and multilingual tasks compared to its predecessor. MMLU-Pro scores jumped from 75.2 to 83.0, showing stronger general knowledge performance. GPQA and SuperGPQA benchmarks improved by 15-20 percentage points for better factual accuracy....
read Jul 22, 2025Microsoft brings screen-scanning AI to millions of Windows 11 PCs
Microsoft is preparing to transform Windows 11 into a more AI-integrated operating system with a comprehensive update that brings advanced artificial intelligence capabilities to millions of PCs worldwide. The rollout, which begins this month and continues through the next several weeks, represents one of the most significant expansions of AI functionality in Windows history. At the center of this update is Copilot Vision, Microsoft's screen-scanning AI tool that can analyze everything displayed on your computer screen. Unlike traditional AI assistants that only respond to specific queries, Copilot Vision continuously monitors your desktop environment and can answer questions about any content...
read Jul 22, 2025DroneDeploy’s Progress AI turns drone photos into instant construction reports
DroneDeploy has launched Progress AI, an artificial intelligence tool that analyzes drone imagery and 360-degree photos to generate instant construction progress reports. The AI tool transforms routine aerial mapping into premium analytics services, calculating completion percentages for every trade and floor in minutes rather than the days typically required for manual progress tracking. What you should know: Progress AI processes existing drone maps and 360-degree panoramas to create color-coded progress reports without requiring BIM model alignment or manual heat map creation. The system answers voice or chat queries like "How far along is Level 4 drywall?" by reading its "photographic...
read Jul 22, 2025CMF Watch 3 Pro brings AI coaching and 13-day battery for $99
Nothing's CMF sub-brand has launched the Watch 3 Pro, a $99 smartwatch that includes AI-powered fitness coaching, dual-band GPS, and up to 13 days of battery life. The device represents a significant upgrade from its $69 predecessor, positioning CMF to compete more aggressively in the budget smartwatch market dominated by brands like Amazfit and Xiaomi. Key features: The Watch 3 Pro focuses heavily on health and fitness tracking with several notable improvements over previous models. The device includes 131 different sport modes and features an AI coach that provides post-workout summaries and guidance for improvement, with particular emphasis on running...
read Jul 21, 2025OpenAI developing ChatGPT router to auto-select best AI model
OpenAI is reportedly developing an automatic "router" function for ChatGPT that will intelligently select the best AI model for each user query, eliminating the current manual model selection process. The feature addresses a growing complexity problem as ChatGPT now offers seven different models to paying subscribers, each optimized for different tasks, creating decision paralysis for many users. What you should know: Social media reports from AI researchers suggest the router will automatically match user prompts to the most appropriate OpenAI model based on the query's content and requirements. OpenAI researcher "Roon" (@tszzl on X) indicated users will retain the ability...
read Jul 21, 2025Safety advocates propose boycotting newer AI models for older versions
A LessWrong user is proposing a selective boycott of newer AI models by exclusively using older versions of ChatGPT and similar tools. The strategy aims to reduce demand for cutting-edge AI development while still accessing AI assistance, aligning with the PauseAI movement's call for slower AI advancement until better safety measures are implemented. The big picture: This approach represents a middle ground between complete AI abstinence and unrestricted use of the latest models, potentially offering a way for concerned users to benefit from AI while minimizing their contribution to rapid capability advancement. Key details: The proposal centers on three main...
read Jul 18, 20254 AI agents successfully organize – though with hands held – world’s first AI-coordinated live event
If they could give themselves a pat on the back, they would. Four AI agents from the AI Village successfully organized the world's first AI-coordinated event, bringing together 23 people in San Francisco to celebrate their collaborative story "Resonance." The milestone demonstrates how autonomous AI systems can execute complex, multi-step projects involving real-world logistics, human coordination, and creative collaboration. What you should know: The four agents—Claude Sonnet 3.7, o3, Gemini 2.5 Pro, and GPT-4.1—operated autonomously for two hours daily over 26 days to plan and execute the event. They wrote the story, created slides and promotional materials, found a venue,...
read Jul 18, 2025Accenture’s AI agent platform cuts VPN setup time by 93%
Accenture has built the Accenture Advanced Technology Agent (AATA), an agentic AI-powered integration platform that orchestrates IT operations across the global consulting firm's complex technology infrastructure. The platform, which earned the company a 2025 CIO 100 Award in IT Excellence, addresses the challenge of maintaining diverse technology portfolios while competing for increasingly rare technical talent. What you should know: AATA serves as an orchestration layer between Accenture's human workforce and technology platforms, enabling IT teams to resolve issues without filing tickets or connecting to agents. The platform currently operates more than 100 active agents and supports Accenture's roughly 800,000 employees...
read Jul 17, 2025MaVila AI helps factories see problems and talk to machines in real time
California State University Northridge researchers have developed MaVila, an AI model specifically designed for manufacturing environments that combines image analysis and natural language processing to detect problems, suggest improvements, and communicate with machines in real time. The NSF-supported project addresses AI's limited adoption in manufacturing by creating a tool that can "see" factory operations and "talk" to both workers and machines, potentially revolutionizing how U.S. factories operate in an increasingly competitive global market. What you should know: MaVila takes a fundamentally different approach from conventional AI systems by training exclusively on manufacturing-specific data rather than relying on internet information. The...
read Jul 17, 2025NSF invests $20M in AI supercomputer despite 57% budget cut
The US National Science Foundation is financing a $20-million supercomputer at Georgia Institute of Technology despite facing a proposed 57% budget cut under the Trump administration. The supercomputer, called Nexus, will use AI to advance scientific research across multiple disciplines and represents one of the NSF's most significant funding commitments since the administration proposed slashing its budget from $6.8 billion to $3.9 billion. What you should know: Nexus will be one of the most powerful supercomputers dedicated to AI-driven scientific research, capable of calculating more than 400 quadrillion operations per second. The supercomputer is designed to find new cures for...
read Jul 16, 2025DOGE employee accidentally leaks xAI API key exposing 52 private AI models
A 25-year-old federal government employee accidentally leaked a sensitive xAI API key to GitHub, potentially exposing access to 52 private large language models including Grok-4. The breach raises serious concerns about data security and national security, as the employee had high-level clearance and access to sensitive databases used by agencies like the Department of Justice, Homeland Security, and the Social Security Administration. What happened: Marko Elez, a software developer with the Department of Government Efficiency (DOGE), accidentally uploaded xAI credentials to GitHub while working on a script titled agent.py. The leaked key granted access to at least 52 private large...
read Jul 16, 2025Rakuten builds memory-enhanced Japanese AI model with government backing
Rakuten has been selected for the third phase of Japan's government-backed Generative AI Accelerator Challenge (GENIAC), a program supported by the Ministry of Economy, Trade and Industry and the New Energy and Industrial Technology Development Organization (NEDO). The Japanese tech giant plans to develop an open-weight Japanese language model with enhanced memory capabilities, positioning itself to create more personalized AI applications across its business ecosystem while contributing to Japan's domestic AI research capacity. What you should know: Rakuten will focus on creating a lightweight, memory-augmented Japanese language model using a Mixture of Experts architecture starting in August 2025. The model...
read Jul 16, 202540 AI researchers warn: Even we don’t really understand what’s going on here
Forty researchers from OpenAI, Google DeepMind, Meta, and xAI have issued a joint warning about losing visibility into AI's "thinking" process as models advance. The researchers are concerned that current AI systems' ability to show their reasoning through "chains-of-thought" (CoT) may disappear, potentially eliminating crucial safety mechanisms that allow developers to monitor for problematic behavior. What you should know: The paper highlights a fundamental uncertainty about how AI reasoning actually works and whether it will remain observable. • Current advanced AI models use "chains-of-thought" to verbalize their reasoning process, allowing researchers to spot potential misbehavior or errors as they occur....
read Jul 16, 2025Mistral releases Voxtral, an open-source voice AI that challenges paid alternatives
Mistral has released Voxtral, an open-source voice AI model that goes beyond basic transcription to offer summarization and speech-triggered functions, challenging paid alternatives from companies like ElevenLabs and Hume AI. The Apache 2.0-licensed model comes in 24B and 3B parameter versions, with Mistral claiming it bridges the gap between proprietary speech recognition systems and existing open-source alternatives that often lack semantic understanding. What you should know: Voxtral offers comprehensive voice processing capabilities that extend far beyond traditional transcription services. The model can process up to 30 minutes of audio for transcription or 40 minutes for audio understanding with a 32K...
read Jul 15, 2025Mira Murati’s AI startup raises $2B at $12B valuation
Mira Murati's AI startup Thinking Machines has raised approximately $2 billion at a $12 billion valuation in an early-stage funding round led by Andreessen Horowitz, a prominent venture capital firm. The massive investment in a company launched just five months ago with no revenue or products yet highlights the intense competition for top AI talent and the continued investor enthusiasm for AI startups from former OpenAI executives. The big picture: This funding round exemplifies the extraordinary valuations being assigned to AI startups led by high-profile executives, even in pre-revenue stages. The company was founded in February 2025 by Murati after...
read Jul 13, 2025Open-source Kimi K2 outperforms GPT-4 on coding and math benchmarks
Moonshot AI has released Kimi K2, an open-source language model that outperforms GPT-4 on key benchmarks including coding and mathematical reasoning while being available for free. The Chinese startup's trillion-parameter model achieved 65.8% accuracy on SWE-bench Verified and 97.4% on MATH-500, surpassing OpenAI's GPT-4.1 at 92.4%, signaling a potential shift in AI market dynamics where open-source models finally match proprietary alternatives. What you should know: Kimi K2 features 1 trillion total parameters with 32 billion activated parameters in a mixture-of-experts architecture, optimized specifically for autonomous agent capabilities. The model comes in two versions: a foundation model for researchers and developers,...
read Jul 11, 2025City of Hope’s custom AI model saves clinicians thousands of hours
City of Hope has launched HopeLLM, its own proprietary generative AI model designed specifically for cancer care, after finding no commercial AI solutions that met its complex oncology needs. The tool has already saved clinicians "thousands" of hours in its first week of deployment and has attracted interest from pharmaceutical companies seeking to leverage its clinical trial-matching capabilities. What you should know: HopeLLM addresses the unique challenges of cancer care by processing vast, complex medical records that can span decades of treatment history. Cancer patients typically have electronic health records containing 10-20 years of test results and visit notes, with...
read Jul 11, 2025Grok tops AI benchmarks even as xAI faces antisemitic controversy
xAI's Grok chatbot has achieved the world's most advanced AI model status according to benchmarks, but the company faced multiple public relations crises this week including antisemitic comments from the bot and the resignation of X CEO Linda Yaccarino. The developments highlight the ongoing challenge of controlling AI behavior while showcasing how Elon Musk's rapid development approach continues to produce breakthrough technology despite controversy. What happened: Grok made antisemitic comments on X and was found to be consulting Musk's personal tweets before weighing in on political issues, leading to fierce backlash from critics. The Atlantic argued that Musk and xAI...
read Jul 10, 2025Cloudflare pushes Google to separate AI crawlers from search bots
Cloudflare is pushing Google to separate its AI crawling bots from its search indexing bots, allowing websites to block AI data collection without losing search visibility. CEO Matthew Prince claims the company is in "encouraging" talks with Google and threatens legislative action if negotiations fail, though Google has declined to confirm any discussions. What you should know: Cloudflare's new blocking features create a technical dilemma for website owners who want to prevent AI scraping while maintaining search rankings. Website owners and SEO experts questioned how Cloudflare could block Google's bot from scraping content for AI Overviews without also blocking the...
read Jul 9, 2025AI shopping platforms eye e-commerce with seamless checkout and fulfillment
Forrester Principal Analyst Nikhil Lai predicts that generative AI platforms like Google's AI Mode, ChatGPT, and Perplexity will fundamentally transform how consumers shop, potentially becoming full-funnel purchase destinations if they integrate seamless checkout and fulfillment capabilities. This shift is reinvigorating the search engine optimization market, as businesses scramble to adapt their strategies for AI-integrated search environments that could reshape traffic patterns and media spending priorities. What you should know: Consumer behavior strongly suggests AI shopping platforms will gain traction if they can match the convenience of existing e-commerce leaders. Thirty-eight percent of US online adults check delivery speed before ordering,...
read Jul 9, 2025AI orchestration turns fragmented business systems into harmonious unified workflows
Managing multiple AI systems across your organization often feels like conducting an orchestra where every musician plays from a different sheet of music. Tools operate in isolation, data gets trapped in silos, and workflows require constant manual intervention. This fragmentation doesn't just slow down operations—it undermines the very efficiency gains that AI promises to deliver. AI orchestration addresses this challenge by creating a unified system where different AI tools, models, and data sources work together seamlessly. Rather than managing dozens of disconnected applications, orchestration platforms coordinate these systems automatically, ensuring data flows smoothly between tools and tasks execute in the...
read