language models

Winsage
July 23, 2025
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.
AppWizard
July 17, 2025
Gemini has begun its phased rollout on Android watches equipped with Wear OS 4 and above. It offers improved natural language command recognition and tailored responses from Google's Search LLM, allowing users to interact seamlessly with Google and Samsung applications. Gemini can create Google Calendar reminders and YouTube Music playlists through voice commands. It provides specific examples in response to queries, such as suggesting foods rich in carotenoids, and can deliver concise answers to spontaneous inquiries. Gemini efficiently pulls data from Search and Maps for tasks like finding local theater showtimes and creating calendar entries. Users can select default applications for tasks, enhancing speed and functionality. Battery consumption is manageable, with only about 20% used during a full afternoon of heavy use. However, it occasionally misinterprets commands and could benefit from brevity in responses. A "View on phone" button and better recognition of speech completion would improve user experience. Support for the new One UI 8 Watch Tile format could enhance functionality further.
AppWizard
July 16, 2025
Recent research indicates a 681% increase in the number of games on Steam that acknowledge the use of generative AI, rising from around 1,000 titles in 2024 to over 7,500. Ichiro Lambe, a game developer, noted that the applications of generative AI in gaming include audio elements, narrative generation, marketing materials, and code generation assistance. The use of AI during runtime is also increasing, with examples such as flagging offensive content and generating 3D worlds based on player prompts. Some games featuring AI have received positive reviews, suggesting growing acceptance among players. Notable titles using generative AI include My Summer Car, which has sold 2.5 million copies. Developers are using carefully crafted language in disclosures about AI to reassure customers, and there are concerns about undisclosed AI usage in some games.
Winsage
July 13, 2025
Microsoft's Copilot has transformed into a conventional AI assistant, disappointing some users who expected seamless integration with Windows. In response, a third-party solution called Windows-MCP has emerged, utilizing advanced language models like Anthropic's Claude to help users manage their PCs. Setting up Windows-MCP requires Python 3.13 or newer, the Claude desktop app, the UV package manager for Python, Node.js, and Git. Once installed, users can execute commands using text, but Claude's permission requirements can disrupt command flow. During testing, Claude successfully opened and closed browsers, launched applications, and summarized web content, though the free plan limits interactions. Users have reported that the time taken for Claude to execute commands can be longer than performing tasks manually, and bugs have been noted, including issues with browser profiles. Despite its current limitations, Windows-MCP showcases the potential for AI to enhance user interaction with technology.
Tech Optimizer
July 10, 2025
EnterpriseDB (EDB) has returned to the Goodwood Festival of Speed, focusing on the impact of advanced AI data systems in the automotive industry. A global survey by EDB of 190 automotive executives revealed that only 13% of organizations are successfully leveraging agentic and generative AI, with these companies achieving a fivefold return on investment through a focus on data sovereignty. The automotive sector is undergoing a significant transformation, with the U.K. automotive AI market projected to grow from £8 million in 2024 to £2 million by 2030, reflecting a 21.7% CAGR. EDB is also fostering future talent through initiatives that encourage students to propose innovative data and AI solutions, with a £2,000 prize for the winning idea. The U.K. STEM workforce is expected to grow by 10% by 2030, while many tech firms face challenges in filling roles. EDB has received multiple recognitions for its innovation and workplace culture, including listings in various industry awards. EDB Postgres® AI is described as an open, enterprise-grade sovereign data and AI platform that integrates various workloads while ensuring compliance and security.
Tech Optimizer
July 9, 2025
Amazon Bedrock Knowledge Bases has introduced a fully managed Retrieval Augmented Generation (RAG) feature that connects large language models (LLMs) with internal data sources, enhancing the relevance and accuracy of responses by integrating contextual information from private datasets. During AWS re:Invent 2024, it was announced that this feature now supports natural language querying for structured data retrieval from Amazon Redshift and Amazon SageMaker Lakehouse, allowing generative AI applications to access both structured and unstructured data sources. The system converts user queries into SQL queries using natural language processing, enabling data retrieval without requiring users to know SQL syntax. Amazon Bedrock Knowledge Bases currently supports structured data retrieval from Amazon Redshift and SageMaker Lakehouse. Although direct support for Aurora PostgreSQL-Compatible is not available, users can utilize zero-ETL integration between Aurora PostgreSQL-Compatible and Amazon Redshift to make their data accessible. This integration replicates Aurora PostgreSQL tables to Amazon Redshift in near real-time, simplifying data management. To enable natural language querying of structured application data stored in Aurora, organizations can set up an Aurora PostgreSQL database, create a schema with interconnected tables (products, customers, and orders), and populate these tables with sample data while maintaining referential integrity. Subsequently, they can establish zero-ETL integration with Amazon Redshift, which involves creating a Redshift Serverless workgroup and mapping the database for synchronization. Once the zero-ETL integration is verified, organizations can create an Amazon Bedrock knowledge base for natural language querying. This requires granting appropriate permissions to the Amazon Bedrock Knowledge Bases AWS Identity and Access Management (IAM) role and ensuring the knowledge base is synchronized with Amazon Redshift. After setting up the knowledge base, users can execute natural language queries, which are translated into SQL and processed to generate human-readable responses. Examples of queries include counting unique customers and identifying customers who have purchased the most products. Finally, it is recommended to clean up resources after use to avoid ongoing charges.
AppWizard
July 8, 2025
Android users will soon experience an updated Photo Picker that includes a search bar for quicker photo selection. This update, which is gradually rolling out, enhances the previous basic gallery interface by allowing users to input keywords to find specific images. The search function does not utilize the same large language models as Google’s “Ask Photos” feature but aims to simplify the process of locating photos among many. The Photo Picker was initially introduced to improve privacy and security for applications accessing images. The update is currently being rolled out slowly, with many devices yet to receive it.
Winsage
June 23, 2025
Mu utilizes a one-time encoding method to separate input and output tokens, reducing computational demands and memory usage. In performance comparisons, Mu showed speed advantages over Microsoft’s Phi-3.5, performing nearly on par with a fine-tuned version of Phi-3.5-mini while being only one-tenth of its size. These enhancements are crucial for real-time processing applications on devices.
Search