News

Google’s new AI model, Gemini 2.0 Flash Thinking Experimental, focuses on reasoning and problem-solving in fields like math, coding, and physics. Unlike typical AI, it fact-checks itself and pauses to explain its reasoning before giving answers. While this improves accuracy, it makes the model slower, sometimes taking seconds to minutes to respond. However, like many… Continue reading Google releases its own ‘reasoning’ AI model

ModernBERT, a new model developed by Answer.AI and LightOn, is a major upgrade to BERT, one of the most widely used AI models since 2018. ModernBERT is faster, more accurate, and can process up to 8,000 tokens compared to BERT’s 512. It also incorporates code in its training data, making it suitable for tasks like… Continue reading ModernBERT: The Next Evolution in Efficient AI Models

An NHS trust in Hull and East Yorkshire is enhancing its Lenus app for chronic obstructive pulmonary disease (COPD) patients by integrating artificial intelligence (AI). The app, launched in March 2023, helps patients track their symptoms and communicate directly with NHS clinics. It has already reduced hospital visits by 40%. With AI, the system will… Continue reading Hospitals to use AI to cut patient visits

The upcoming o1 model on Microsoft Azure OpenAI Service brings powerful AI capabilities with support for text and image inputs, making it ideal for complex tasks like problem-solving and data analysis. It features advanced reasoning tools, a 200K token context window for handling long inputs, and structured outputs for better control. Businesses like Rohirrim, Wrtn… Continue reading Announcing the o1 model in Azure OpenAI Service

NVIDIA’s NeMo Retriever helps businesses work with information in multiple languages, making AI systems smarter and more efficient. It can search and process data across languages, connect to large databases, and handle long documents like contracts or medical records. By using advanced techniques, it reduces storage needs by 35x, making AI solutions faster, cheaper, and… Continue reading NVIDIA Debuts NeMo Retriever Microservices for Multilingual Generative AI Fueled by Data

OpenAI has rolled out ChatGPT Search for all users, allowing it to fetch real-time web information like news, sports scores, and business details. Users can access it via a “Search the web” button or let ChatGPT decide when to use it. It provides summaries, links, and multimedia like videos and images. Previously available only to… Continue reading OpenAI brings its AI-powered web search tool to more ChatGPT users

Google DeepMind has introduced Veo 2, a new video-generation AI capable of creating detailed, high-resolution videos up to 4K and over two minutes long. While still limited in its current tool, VideoFX, Veo 2 improves on motion, lighting, and camera controls, producing sharper, more realistic videos. However, challenges like consistency and detail remain. The model… Continue reading Google DeepMind Unveils Veo 2 and Imagen 3

Meta has unveiled new AI innovations, including Meta Motivo, which enhances digital avatar movements for realistic Metaverse interactions. Another model, the Large Concept Model (LCM), simplifies reasoning by focusing on high-level ideas rather than individual words, making it efficient and multilingual. Meta also introduced Meta CLIP 1.2, a vision-language encoder, to improve AI’s ability to… Continue reading Meta Advances AI with Motivo, LCM, and Meta CLIP

NVIDIA is transforming robotics development with its Isaac Sim platform, allowing robots to be trained and tested in virtual environments before deployment. Built on NVIDIA Omniverse, Isaac Sim helps developers simulate realistic scenarios using synthetic data, reducing costs and time while enhancing safety. It’s now available on powerful Amazon EC2 G6e instances for better performance.… Continue reading Into the Omniverse: How OpenUSD-Based Simulation and Synthetic Data Generation Advance Robot Learning