Google's Nested Learning aims to stop LLMs from catastrophic forgetting

Google is planning a significant expansion of its AI infrastructure, aiming to increase its computing capacity by 1,000 times over the next five years. This ambitious goal reflects the company's response to the surging demand for artificial intelligence capabilities, as outlined in internal communications from its AI infrastructure chief.

Researchers at TU Darmstadt have introduced the VOIX framework, which adds two new HTML elements to websites, enabling AI agents to recognize available actions without needing to interpret complex user interfaces visually. This innovation aims to enhance the interaction between AI and web environments.

Meta has unveiled the third generation of its Segment Anything Model (SAM 3), which utilizes an open vocabulary to enhance its understanding of images and videos. This model distinguishes itself from traditional segmentation models by employing a novel training method that integrates both human and AI annotators.

OpenAI is under pressure as an internal memo reveals CEO Sam Altman's response to Google's advancements with its Gemini 3 model. The memo outlines OpenAI's strategy, codenamed 'Shallotpeat', to regain competitive ground in the AI landscape.

OpenAI's GPT-5 Science Acceleration report highlights how researchers are utilizing the model to streamline their daily tasks. The report provides insights into the practical applications of AI in scientific research while emphasizing the continued need for human oversight in decision-making processes.

OpenAI has launched a group chat feature in ChatGPT, initially testing it in Japan, South Korea, Taiwan, and New Zealand. This new functionality aims to enhance user interaction by allowing multiple users to engage in conversations simultaneously, facilitating collaboration and dynamic discussions.

Google has introduced an updated image model named Nano Banana Pro, also referred to as Gemini 3 Pro Image. This model aims to enhance the intentionality of image generation, providing users with improved capabilities for creating realistic AI-generated images.

The Allen Institute for AI (Ai2) has introduced OLMo 3, the first fully open AI model designed to expose its reasoning process. This new 32B 'thinking' model operates 2.5 times more efficiently than comparable models, marking a significant advancement in AI transparency and functionality.