Latest AI News

Federal Reserve Governor Lisa D. Cook revealed that while the Federal Open Market Committee (FOMC) isn't using artificial intelligence for policy decisions, the Fed is actively deploying AI tools to enhance writing, coding, and research capabilities. In her July 17 speech at the National Bureau of Economic Research in Cambridge, Cook emphasized that AI is transforming the economy by accelerating innovation and worker efficiency, potentially affecting both employment and price stability. The Fed is carefully studying these economic implications while simultaneously experimenting with AI applications internally.

Amazon Web Services has developed the In-Row Heat Exchanger (IRHX), a hybrid liquid-air cooling system designed specifically for Nvidia's power-hungry Blackwell GPUs used in advanced AI workloads. The innovative solution addresses the extreme heat generated by high-density GPU racks without requiring major data center renovations or increasing water consumption. This technology enables AWS to offer its new P6e instances featuring Nvidia's GB200 NVL72 platform, which packs 72 interconnected GPUs into a single rack for unprecedented AI computing power.

MIT researchers have developed CodeSteer, an intelligent assistant that guides large language models to switch between text and code generation until correctly answering complex queries. The system increased LLM accuracy on symbolic tasks like math problems and spatial reasoning by more than 30%, enabling less sophisticated models to outperform more advanced ones. This breakthrough could significantly improve AI problem-solving capabilities for complex tasks in robotics, supply chain management, and other fields requiring precise computational reasoning.

A comprehensive study led by MIT researchers has identified key challenges preventing AI from fully automating software development. Published on July 16, 2025, the research led by Professor Armando Solar-Lezama outlines a roadmap for advancing beyond simple code generation to tackle complex engineering tasks. The study calls for community-scale efforts to develop better benchmarks, improve human-AI collaboration, and create richer datasets that capture real development processes.

OpenAI is preparing to integrate a checkout system into ChatGPT that would allow the company to earn commissions on e-commerce sales made through the platform. This strategic move, reported on July 17, 2025, builds upon OpenAI's April partnership with Shopify and aims to monetize the platform's massive free user base. The initiative represents a significant shift in OpenAI's revenue model as it seeks new income streams beyond its subscription services.

OpenAI CEO Sam Altman has confirmed that GPT-5, the company's next major AI model, is scheduled for release in summer 2025. Early testers describe the upcoming model as 'materially better' than its predecessor GPT-4, with significant improvements in performance and capabilities. While discussing potential monetization strategies, Altman indicated he's 'not totally against' ads in ChatGPT but warned that altering model outputs for advertisers would destroy user trust.

BrightAI, a startup applying artificial intelligence to physical infrastructure, has raised $51 million in Series A funding led by Khosla Ventures and Inspired Capital. The company's Stateful platform uses sensors, drones, and edge AI to monitor and maintain critical systems like water pipelines, power grids, and HVAC networks. With this investment, BrightAI plans to expand its team and open a new San Francisco headquarters as it scales its technology across essential industries.

San Diego-based Firestorm Labs has raised $47 million in Series A funding to advance its AI-driven drone manufacturing technology. The investment, led by New Enterprise Associates with participation from Lockheed Martin Ventures and others, will accelerate the company's innovative xCell system—a deployable factory-in-a-box that produces customizable unmanned aerial systems. This funding follows Firestorm's recent $100 million Air Force contract and strategic partnership with HP for mobile 3D printing technologies.

San Francisco-based Confident Security has emerged from stealth with $4.2 million in seed funding to address one of AI's biggest adoption barriers: data privacy. The company's CONFSEC technology, inspired by Apple's Private Cloud Compute architecture, creates an encrypted wrapper around AI models that prevents sensitive data from being stored, viewed, or used for training by model providers or third parties. This innovation could accelerate AI adoption in highly regulated sectors like healthcare, finance, and legal services where privacy concerns have limited implementation.

San Francisco-based Unify has raised $40 million in Series B funding led by Battery Ventures, with participation from OpenAI Startup Fund and other investors, to accelerate its AI-powered sales platform. The company's technology combines real-time buyer intent signals with AI agents to help sales teams identify high-potential prospects and personalize outreach at scale. Founded in 2023, Unify has experienced explosive growth with revenue increasing 8x over the past year, as clients like Perplexity and Airwallex use the platform to generate millions in sales pipeline.

iCOUNTER, led by former Mandiant president John Watters, has emerged from stealth with $30 million in Series A funding to combat increasingly sophisticated AI-enabled cyber threats. The Dallas-based company has developed specialized technology that provides precision risk intelligence focused exclusively on targeted operations against specific organizations and their ecosystems. In an era where traditional threat intelligence is becoming obsolete against novel AI-generated attacks, iCOUNTER aims to help organizations detect and counter threats before they become 'Patient Zero'.

Google DeepMind unveiled AlphaGenome on June 25, 2025, a revolutionary AI system that interprets the human genome's non-coding regions—the 98% of DNA that doesn't produce proteins but regulates gene activity. The model can analyze DNA sequences up to 1 million base-pairs long and predict how genetic variants affect biological processes across multiple cell types. Scientists have praised AlphaGenome for its unprecedented ability to identify how non-coding mutations contribute to diseases like cancer, potentially accelerating treatment development.

Google has expanded its Gemini 2.5 family by making Gemini 2.5 Flash and Pro generally available, while introducing 2.5 Flash-Lite in preview – their most cost-efficient and fastest 2.5 model yet. Flash-Lite is a reasoning model optimized for cost and speed with "thinking" turned off by default, providing an even lower cost option for latency-sensitive use cases that require less model intelligence. Meanwhile, Gemini 2.5 Pro has become the world-leading model across the WebDev Arena and LMArena leaderboards, demonstrating Google's continued AI advancement.

Amazon Web Services has launched Kiro AI, a revolutionary specification-driven agentic integrated development environment that bridges the gap between rapid AI prototyping and production-ready software. Announced on July 14, 2025, Kiro introduces a structured approach to software development by transforming developer prompts into detailed specifications, design documents, and task lists before generating code. This new tool represents Amazon's strategic entry into the competitive AI-powered IDE market, potentially transforming how developers build and maintain software.

Swiss researchers at the Paul Scherrer Institute have developed an AI system that can design low-carbon cement formulations in seconds rather than months. Led by mathematician Romana Boiger, the system simulates thousands of ingredient combinations to identify recipes that maintain structural strength while significantly reducing carbon emissions. With cement production responsible for approximately 8% of global CO2 emissions, this innovation could transform the construction industry's environmental impact.

Researchers from the University of Cambridge and UCL have developed a revolutionary robotic skin technology that enables machines to feel heat, pain, and pressure like humans. The flexible, low-cost gel material transforms a robot's entire surface into a sensitive touch interface, detecting signals through over 860,000 tiny pathways. Unlike traditional approaches requiring multiple sensor types, this single-material solution simplifies manufacturing while enhancing robots' ability to interact safely with their environment.

Google DeepMind has unveiled AlphaGenome, a groundbreaking AI system that predicts how genetic changes affect gene expression in the non-coding regions of DNA. This transformer-based model can analyze up to one million DNA letters simultaneously, helping researchers pinpoint disease causes by tracking genetic mutations' consequences. Available for non-commercial research, AlphaGenome represents a significant advancement in understanding the genome's 'dark matter' that makes up 98% of human DNA.

Amazon Web Services unveiled a comprehensive suite of agentic AI capabilities at the AWS Summit on July 16, 2025, designed to automate complex business processes across applications. The centerpiece, Amazon Bedrock AgentCore, enables organizations to deploy AI agents that can reason, plan, and execute tasks with minimal human oversight. These advancements aim to bridge the gap between AI prototypes and production-ready systems while reducing operational overhead.

European researchers have demonstrated a revolutionary computing method using laser pulses through ultra-thin glass fibers to perform AI computations thousands of times faster than traditional electronics. The breakthrough, led by teams from Tampere University and Université Marie et Louis Pasteur, leverages nonlinear light interactions in optical fibers to create an Extreme Learning Machine architecture that could dramatically reduce energy consumption while increasing processing speeds for AI applications.