Nov 19, 2025

• Gemini 3 is Google's most advanced AI model, enhancing reasoning and multimodal capabilities, enabling users to bring any idea to life across various Google products like the Gemini app and Vertex AI.

• MAKER is the first system to solve a task with over one million LLM steps without errors, showcasing a breakthrough in long-range task execution.

• Gemini 3 Pro enhances capabilities from Gemini 2.5, featuring 1 million input tokens and 64,000 output tokens, with improved performance in multimodal tasks including audio transcription and image processing.

• MMaDA-Parallel introduces a parallel multimodal diffusion framework that enhances text-image generation by allowing continuous interaction between modalities, addressing performance degradation in complex tasks due to error propagation.

• SAM 3 introduces advanced features in AI technology, enhancing user interaction and experience through improved algorithms and user interface design. For more details, visit the official article.

• Mojo-V introduces a privacy-oriented RISC-V extension that enables secure, efficient secret computation by utilizing dedicated secret registers and third-party key encryption, achieving 5-7 orders of magnitude performance improvement over fully homomorphic encryption (FHE).

• Segment Anything Model 3 (SAM 3) introduces a unified framework for detecting, segmenting, and tracking objects in images and videos using concept prompts, significantly enhancing Promptable Concept Segmentation (PCS) with a dataset of 4M unique labels.

• Researchers from the University of Vienna uncovered a significant privacy vulnerability in WhatsApp's contact discovery mechanism, enabling the enumeration of 3.5 billion accounts across 245 countries, which has since been mitigated by Meta.

• Microsoft, NVIDIA, and Anthropic have formed a strategic partnership to scale Anthropic's Claude AI model on Azure, with Anthropic committing to purchase $30 billion in Azure compute capacity and up to 1 gigawatt of additional capacity.

• DeepClause is a neurosymbolic AI system that integrates symbolic reasoning with neural language models, enabling agents to perform complex logic and multi-step reasoning through its Prolog-based DSL, DML (DeepClause Meta Language).

• Reproducible baselines for human action recognition achieve 87.05% accuracy on UCF-101 and 88.5% on Stanford40, providing a reliable starting point for researchers with pretrained models available on HuggingFace.

• NVIDIA and Microsoft are enhancing AI capabilities by integrating next-generation technologies, including the NVIDIA Spectrum-X Ethernet switches and RTX PRO 6000 Blackwell GPUs, to power the new Microsoft Fairwater AI superfactory for large-scale AI training and inference.

• Cornserve is an open-source platform designed to serve complex any-to-any multimodal AI models, such as Qwen 3 Omni, by utilizing a microservices architecture that enhances flexibility and scalability.