News from the AI & ML world
DeeperML - #aicopilot
|
Jibin Joseph@PCMag Middle East ai
//
Microsoft is expanding its AI capabilities by testing a new Copilot Live Portraits feature and making OpenAI's Sora video generator accessible for free through Bing Video Creator. The Copilot Live Portraits are currently in an experimental phase and introduce a new interface element with customizable avatars for users. These avatars, offering a selection of visual styles for male and female figures, could potentially serve as the visual interface for Copilot in voice-based interactions, creating a more human-like experience. Internal references suggest Microsoft might be developing real-time, visually expressive characters, aligning with the broader trend of synthetic video avatars in the AI space. The integration of Live Portraits may also influence the future of Copilot Characters, possibly merging both into a spectrum of assistants ranging from fixed personas to customizable 3D portraits.
Microsoft has launched Bing Video Creator with Sora integration for mobile users. Powered by OpenAI's Sora, the new tool transforms text prompts into short videos, offering users a free way to bring their creative ideas to life. The Bing Video Creator is available on the Bing mobile app for iOS and Android, allowing users to generate short video clips by simply describing what they want to see. This initiative follows the release of Bing Image Creator and Copilot, expanding Microsoft's AI-driven offerings. Bing Video Creator generates five-second videos in a 9:16 format, with plans to support 16:9 format in the future. The service operates on a two-tier speed system, with standard generation being free for all users and fast generation requiring Microsoft Rewards points after an initial allocation of ten free fast creations. Videos are stored for 90 days, and the platform supports direct sharing via email, social media, or generated direct links. Microsoft will implement the safeguards used by OpenAI for Sora, blocking potentially harmful prompts and watermarking outputs based on the C2PA standard. Recommended read:
References :
@www.microsoft.com
//
Microsoft is aggressively pursuing AI integration across a range of its products and research initiatives. One notable project is Aurora, an AI model developed by Microsoft Research to revolutionize environmental forecasting. Senior Researchers Megan Stanley and Wessel Bruinsma are leading the charge, and recently discussed how Aurora extends its capabilities to predict tropical cyclones and ocean waves. Their work, detailed in a Nature publication titled "A Foundation Model for the Earth System," highlights Aurora's potential to significantly improve weather prediction accuracy and environmental understanding.
Microsoft is also experimenting with "AI Developing AI" through Project Amelie. This experimental AI agent aims to automate the creation of machine learning pipelines from a single prompt. Microsoft reports that in early testing, Project Amelie outperforms current state-of-the-art benchmarks on MLE-Bench, a framework that evaluates how effectively machine learning agents handle real-world tasks. During a live demo at Microsoft Build 2025, Project Amelie was showcased and how it can assist data scientists by automating the process of data sourcing, analysis, and formatting. This addresses the typically manual and time-consuming tasks involved in preparing data for machine learning. However, Microsoft's AI endeavors are not without their challenges. Research by Alex Lu, a senior researcher at Microsoft, indicates that zero-shot foundation models in single-cell biology currently underperform compared to simpler methods. Lu's work suggests that more research is needed to rigorously evaluate and improve the efficacy of AI models in biological applications. These findings, highlighted in the paper "Assessing the Limits of Zero-Shot Foundation Models in Single-Cell Biology," emphasize the importance of continued research and development to fully realize the potential of AI in scientific domains. Recommended read:
References :
Thomas Claburn@The Register
//
Microsoft is significantly advancing human-agent collaboration with the latest upgrades to its Microsoft 365 Copilot. The tech giant is rolling out new updates, including Researcher and Analyst agents designed to enhance workplace productivity by providing in-depth research and data analysis. These AI agents, powered by OpenAI's deep reasoning models, are envisioned as digital colleagues capable of performing complex workplace tasks, helping professionals tackle intricate challenges with advanced reasoning capabilities. This aligns with Microsoft’s broader AI-first strategy, aiming to scale digital labor and drive substantial productivity gains for companies adopting these technologies.
Microsoft is also unveiling a redesigned Microsoft 365 Copilot app featuring AI-powered search and an Agent Store, positioning it as the central hub for human-agent collaboration. The new AI-powered enterprise search tool, Copilot Search, organizes data across the enterprise, providing rich, context-aware answers from first-party and third-party apps like ServiceNow, Google Drive, and Jira. The Agent Store provides access to Microsoft’s Researcher and Analyst agents, initially introduced in March and available to those enrolled in Microsoft’s Frontier program. Moreover, Microsoft is adding new capabilities to its Control System feature to assist IT professionals in overseeing and measuring bot usage effectively. These updates are part of a broader effort to integrate AI across the Microsoft ecosystem, addressing the increasing demand for AI-powered solutions in the enterprise. According to Microsoft’s 2025 Work Trend Index report, a significant majority of companies are rethinking their strategies to leverage AI, signaling a decisive move toward full-scale AI transformation. Key features include AI-powered enterprise search, personalized memory capabilities, specialized reasoning agents like Researcher and Analyst, and the Agent Store. The Researcher agent helps with multi-step research tasks, while the Analyst agent provides data science capabilities. Recommended read:
References :
Megan Crouse@techrepublic.com
//
Microsoft has unveiled BitNet b1.58, a groundbreaking language model designed for ultra-efficient operation. Unlike traditional language models that rely on 16- or 32-bit floating-point numbers, BitNet utilizes a mere 1.58 bits per weight. This innovative approach significantly reduces memory requirements and energy consumption, enabling the deployment of powerful AI on devices with limited resources. The model is based on the standard transformer architecture, but incorporates modifications aimed at efficiency, such as BitLinear layers and 8-bit activation functions.
The BitNet b1.58 2B4T model contains two billion parameters and was trained on a massive dataset of four trillion tokens, roughly equivalent to the contents of 33 million books. Despite its reduced precision, BitNet reportedly performs comparably to models that are two to three times larger. In benchmark tests, it outperformed other compact models and performed competitively with significantly larger and less efficient systems. Its memory footprint is just 400MB, making it suitable for deployment on laptops or in cloud environments. Microsoft has released dedicated inference tools for both GPU and CPU execution, including a lightweight C++ version, to facilitate adoption. The model is available on Hugging Face. Future development plans include expanding the model to support longer texts, additional languages, and multimodal inputs such as images. Microsoft is also working on another efficient model family under the Phi series. The company demonstrated that this model can run on a Apple M2 chip. Recommended read:
References :
Megan Crouse@techrepublic.com
//
Microsoft has unveiled BitNet b1.58 2B4T, a groundbreaking AI model designed for exceptional efficiency. Developed by Microsoft's General Artificial Intelligence group, this model utilizes one-bit neural network weights, representing each weight with only three discrete values (-1, 0, or +1). This approach, called ternary quantization, allows each weight to be stored in just 1.58 bits, drastically reducing memory usage. The result is an AI model that can operate on standard CPUs without the need for specialized, energy-intensive GPUs.
Unlike conventional AI models that rely on 16- or 32-bit floating-point numbers, BitNet's unique architecture allows it to run smoothly on hardware like Apple's M2 chip, requiring only 400MB of memory. To compensate for its low-precision weights, BitNet b1.58 2B4T was trained on a massive dataset of four trillion tokens, the equivalent of approximately 33 million books. This extensive training enables the model to perform on par with, and in some cases even better than, other leading models of similar size, such as Meta's Llama 3.2 1B, Google's Gemma 3 1B, and Alibaba's Qwen 2.5 1.5B. To facilitate the deployment and adoption of this innovative model, Microsoft has released a custom software framework called bitnet.cpp, optimized to take full advantage of BitNet's ternary weights. This framework is available for both GPU and CPU execution, including a lightweight C++ version. The model has demonstrated strong performance across a variety of tasks including math and common sense reasoning in benchmark tests. Microsoft plans to expand BitNet to support longer texts, additional languages, and multimodal inputs like images, while also working on the Phi series, another family of efficient AI models. Recommended read:
References :
|
BenchmarksBlogsResearch Tools |