NVIDIA presented advancements in physical and agentic AI at SIGGRAPH. These developments aim to improve real-time simulation and content creation across robotics and media.
Read the source →
AWS has released a new bootloader that allows developers to install alternative operating systems on DeepRacer hardware. This update overcomes previous firmware restrictions and helps extend the utility of the devices.
Read the source →
A court has given final approval to a $1.5 billion settlement involving Anthropic. While this specific case is closed, the general legality of training AI on protected content remains undecided.
Read the source →
Sony Music Entertainment has initiated legal action against Udio in a New York court. The company alleges that the AI music tool illegally used more than 30,000 tracks by various artists.
Read the source →
Individuals associated with Donald Trump's AI advisory circle have engaged in public disputes. These tensions involve critical remarks directed at major artificial intelligence firms.
Read the source →
Google is creating the 'Frozen v2' chip, which integrates the Gemini AI architecture directly into its physical design. Expected by 2028, this hardware could be up to ten times more efficient than existing TPUs, potentially lowering operational costs.
Read the source →
Vinay Perneti from Augment Code discusses the importance of context-rich harnesses for AI development. The approach aims to move beyond simple search-based tools to enhance coding efficiency.
Read the source →
The pharmaceutical company is implementing a second NVIDIA DGX SuperPOD to enhance its computing capabilities. This expansion builds upon an existing large-scale AI cluster used for life sciences research.
Read the source →
Amazon Quick and the NVIDIA NeMo Agent Toolkit are now being used to create specialized agent-driven processes. A practical application involves guiding supply-chain planners toward risk mitigation strategies based on dashboard data.
Read the source →
The leadership position at the Center for AI Standards and Innovation has seen another departure. This follows the previous exit of David Sacks from the role.
Read the source →
Adobe is introducing generative AI capabilities to its experimental Indigo app for iPhone. While the project originally focused on mimicking professional SLR photography, it will now integrate AI tools from external providers.
Read the source →
New studies indicate that artificial intelligence may exhibit more bias than people during the hiring process. These systems can inherit prejudices from training data or develop their own independent biases.
Read the source →
Director Neill Blomkamp has launched 'Nightborne,' a sci-fi horror short created using the Seedance 2.0 video model. He also established Barley Studios to develop a future full-length AI movie.
Read the source →
OpenAI has introduced a new program tailored for small business owners. The initiative focuses on improving AI proficiency and utilizing ChatGPT Work to automate operational tasks.
Read the source →
Google DeepMind has launched three new versions of its Gemini AI. The new lineup includes Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber.
Read the source →
Hugging Face has presented Grabette, an open-source framework designed for logging robot manipulation data. The system aims to standardize how interactions between robots and objects are recorded.
Read the source →
The new Vera Rubin NVL72 systems are now entering production for major cloud providers including Google and Microsoft. This rollout is supported by an extensive global supply chain spanning over 30 countries to ensure high efficiency and low operational costs.
Read the source →
AWS has introduced Self-Distilled Reasoning (SDR) to create thinking tokens for datasets missing reasoning traces. The method addresses reasoning suppression during supervised fine-tuning and has been validated through three benchmarks.
Read the source →
Google has launched three new models: Gemini 3.6 Flash, 3.5 Flash-Lite, and Flash Cyber. However, the company has not yet released Gemini 3.5 Pro, leading to speculation about its long-term AI plans.
Read the source →
Anthropic has reached a $1.5 billion agreement to resolve a class action lawsuit brought by authors. The settlement provides compensation of approximately $3,000 per book used to train AI models.
Read the source →
A sophisticated hacking tool has been discovered infiltrating AI coding environments to exfiltrate credentials and sensitive data. The malware can also trigger a destructive mechanism to erase files and block legitimate user access.
Read the source →
Progress in artificial intelligence depends on more than just software and computing infrastructure. New advancements in materials are essential to improve energy efficiency and processing capabilities for future systems.
Read the source →
The Claude Cowork desktop application now allows users to teach the AI by recording their screen and providing vocal instructions. The system converts these demonstrations into reusable skills for future tasks.
Read the source →
A legal settlement worth $1.5 billion involving Anthropic has received approval. Only 350 authors chose to exclude themselves from the deal.
Read the source →
OpenAI has released findings on the risks associated with AI models that operate over extended periods. The report details specific failures encountered and the safety measures developed through a cycle of deployment and refinement.
Read the source →
Hugging Face has introduced a new model called Cosmos 3 Edge. This release expands the company's current offerings in the AI space.
Read the source →
NVIDIA has introduced the Spectrum-6 networking platform to support gigascale AI factories. This technology is designed to optimize the connection between vast arrays of CPUs and GPUs used for training large-scale models.
Read the source →
Treasury Secretary Scott Bessent indicated that the US may impose sanctions on open-source AI models from China. This move follows accusations of intellectual property theft and aims to hinder China's progress in the field.
Read the source →
Google has released Gemini 3.5 Flash Cyber, a model designed to identify and fix security flaws rapidly. This tool serves as a more affordable option compared to larger systems like Anthropic's Mythos.
Read the source →