language model Archives

Winsage

July 23, 2025

Copilot Vision on Windows 11 sends data to Microsoft servers

Microsoft is integrating artificial intelligence into Windows 11 with new features, including Copilot Vision, which serves as a remote assistant processing information in the background. Windows 11 is positioned as the home for AI, offering features like Recall, Click to Do, and enhanced search capabilities. Copilot Vision captures continuous screenshots for analysis using optical character recognition and a large language model, with data processed on Microsoft servers. User data is not stored long-term, except for conversation transcripts with the Copilot assistant. The update introduces an "agentic" AI capable of adjusting system settings based on natural language commands, currently available on Copilot+ systems with Qualcomm Snapdragon hardware, with Intel and AMD support expected. Microsoft has also enhanced the "Click to Do" feature and introduced AI capabilities in applications like Photos, Paint, and the Snipping Tool. The traditional Blue Screen of Death has been replaced with a Black Screen of Death, which promises a more readable error message format. Additionally, Microsoft has unveiled the Surface Laptop 5G, featuring a 13.8" display, Intel Core Ultra processors with a neural co-processor, and optional 5G connectivity, expected to be available in 2025.

Winsage

July 12, 2025

A new Windows feature could allow your PC to fix itself

Microsoft is introducing a feature called "quick machine recovery" for Windows users, currently available in Windows 11 Build 26100.4762. This feature allows PCs to autonomously troubleshoot and fix issues after a crash, such as a Blue Screen of Death (BSOD), by booting into the Windows Recovery Environment (Windows RE). Quick machine recovery can connect to Microsoft’s servers to send crash data for analysis, eliminating the need for users to interpret error codes. It employs cloud remediation and auto remediation methods, though auto remediation is disabled by default on home PCs. The existing Startup Repair tool will still be available as a fallback option, and users can disable quick machine recovery if desired. Additionally, the build includes a refined "semantic search" capability in Settings, integration of Microsoft’s Reading Coach app, and the ability to export Recall snapshots to third-party applications in Europe.

Tech Optimizer

June 30, 2025

Supercharging AWS database development with AWS MCP servers

The Model Context Protocol (MCP) facilitates integration between AI systems and various data sources, including databases, APIs, and business applications. AWS has released open-source MCP servers for Amazon Aurora, Amazon DynamoDB, and Amazon ElastiCache to enhance developer productivity by allowing AI-assisted tools to access structured metadata. MCP standardizes connections between AI assistants and external environments, enabling intelligent suggestions and context-aware query assistance. Traditional development challenges include time spent on boilerplate queries and context switching between different database technologies, which MCP aims to alleviate. The MCP servers provide secure access to structured metadata and streamline the development process by integrating database context directly into development environments. Developers can utilize these servers to generate tests based on live metadata, monitor database performance, and troubleshoot issues efficiently. The MCP servers are available for local deployment using Docker and can be configured for various AWS database services.

Winsage

June 25, 2025

Microsoft Just Launched Its New Mu Language Model. Should You Buy MSFT Stock Here?

Users can switch the Market flag to access specific data from their country. The platform provides various chart options accessible via the Interactive Chart menu, allowing for customization and analysis. Navigation through symbols is facilitated by up and down arrows for easy movement between data points.

Winsage

June 25, 2025

Microsoft Launches “Mu,” New On-Device AI Model for Copilot+ PCs

Microsoft Corporation unveiled a small language model named Mu on June 23, designed for local operation on personal computers, particularly Copilot+ PCs. Mu is a 330 million parameter encoder-decoder language model optimized for deployment on Neural Processing Units (NPUs), allowing for quick responses with reduced power consumption and memory usage. It was pre-trained on a dataset of hundreds of billions of educational tokens and employs advanced fine-tuning techniques, including distillation and low-rank adaptation, along with transformer enhancements like Dual LayerNorm, Rotary Positional Embeddings, and Grouped-Query Attention. Users will access Mu's capabilities through the Windows operating system's Settings function, translating natural language inputs into system commands.

Winsage

June 25, 2025

Here’s the secret behind Microsoft’s AI agents — meet Mu, the compact AI model transforming Settings on your PC

Microsoft has introduced a small language model named "Mu" to enhance AI agents in the Windows 11 Settings app. This model operates on Neural Processing Units (NPUs) and is designed to automate tasks on users' PCs with granted permissions. Mu was developed with attention to NPU memory constraints, ensuring efficient operation through NPU-optimized processes. The model aims for ultra-low latency, achieving response times under 500 milliseconds, and is integrated into the search box of the Settings app. Currently, AI agents powered by Mu are being tested among Windows Insiders, who can also explore an update to the Recall app.

Winsage

June 24, 2025

Microsoft (MSFT) Launches a New Small Language Model Called Mu

Microsoft has introduced a small language model named Mu, designed to perform complex language tasks directly on devices like Copilot+ PCs, operating on a device's Neural Processing Unit. This model improves response times and reduces power and memory consumption compared to larger cloud-based AI models. Mu is developed using insights from Microsoft's Phi models and high-quality educational data, employing techniques such as distillation and low-rank adaptation for fine-tuning. Mu features an encoder–decoder architecture that separates input and output processes, enhancing speed and minimizing latency. It incorporates innovations like rotary positional embeddings, grouped-query attention, and dual LayerNorm. Microsoft also utilized quantization to reduce memory usage and improve speed without sacrificing accuracy, making Mu ideal for on-device applications. To enhance the AI agent in Windows Settings, Mu was fine-tuned with over 3.6 million examples, improving its ability to understand and manage system settings. The model addresses ambiguous commands and provides rapid assistance with precise multi-word commands, while maintaining a compact size that is one-tenth that of similar AI tools. Analysts on Wall Street have a positive outlook on MSFT stock, giving it a Strong Buy consensus rating based on 30 Buys and five Holds, with an average price target of 6.14 per share, indicating a potential upside of 6%.

Winsage

June 24, 2025

The Settings Agent in Windows 11 Has Its Own AI Model

Microsoft has introduced Mu, a small language model (SML) designed for Copilot+ PCs, which translates natural language queries into function calls within the Windows 11 Settings application. Mu is engineered to efficiently handle complex input-output relationships and operates locally with a response rate exceeding 100 tokens per second. It is fully integrated with the Neural Processing Unit (NPU) and is optimized for performance and efficiency, utilizing a transformer encoder–decoder architecture. Mu evolves from Microsoft’s Phi models, offering comparable performance at one-tenth the size. The model aims to create an AI-powered agent within Settings that understands natural language and adjusts relevant settings seamlessly, ensuring ultra-low latency for a smooth user experience.

Winsage

June 23, 2025

Introducing Mu language model and how it enabled the agent in Windows Settings

Mu is a small language model designed to run locally on Copilot+ PCs, specifically optimized for Neural Processing Units (NPUs). It is a 330M encoder-decoder model that translates natural language queries into actionable settings function calls, enhancing user experience. Mu achieved approximately 47% lower first-token latency and 4.7× higher decoding speed compared to a decoder-only model on a Qualcomm Hexagon NPU. The model incorporates advanced techniques such as Dual LayerNorm, Rotary Positional Embeddings, and Grouped-Query Attention to improve performance while maintaining a compact size. It was trained using A100 GPUs on Azure, pre-training on hundreds of billions of educational tokens and fine-tuning on tasks like SQUAD, CodeXGlue, and the Windows Settings agent. Mu's quantization techniques converted model weights from floating-point to integer representations, optimizing performance on edge devices. The Settings agent was fine-tuned with a dataset of 3.6 million samples to achieve response times under 500 milliseconds, effectively managing user queries related to Windows settings.

AppWizard

May 28, 2025

Messaging app Signal will block Windows Recall from taking chat screenshots

Signal has introduced a "Screen Security" feature in its desktop application to prevent external programs, including Windows 11, from capturing screenshots of its interface. This feature is enabled by default and is a response to Microsoft's Recall feature, which captures app window screenshots for AI processing. Signal is concerned about the privacy risks posed by Recall's design, despite Microsoft's assurances of improved privacy controls. The developers have used Windows' Digital Rights Management technology to protect against unauthorized screenshots due to the lack of official APIs from Windows 11. Signal's leadership has criticized the need for such workarounds and has called on Microsoft to provide better tools for developers to safeguard user privacy.