News

NVIDIA’s MONAI platform is revolutionizing medical imaging by helping doctors and scientists create AI tools for faster and more accurate analysis of X-rays, CT scans, and MRIs. At RSNA’s annual meeting, NVIDIA announced Siemens Healthineers has adopted MONAI Deploy to streamline AI integration in clinical settings, cutting deployment time from months to minutes. MONAI simplifies… Continue reading Siemens Healthineers Adopts MONAI Deploy for Medical Imaging AI

Google has launched Gemini 2.0, a new family of advanced AI models. The chat version, called Gemini 2.0 Flash, is now available globally, while a more experimental multimodal version (with text-to-speech and image generation) is open for developers. The updated model excels at generating code, giving fact-based answers, and performing tasks more efficiently than its… Continue reading Google releases the first of its Gemini 2.0 AI models

Google’s new vision-language model, PaliGemma 2, upgrades the powerful SigLIP image encoder and Gemma 2 text decoder to enhance AI capabilities for understanding and generating image-based text. Available in three sizes (3B, 10B, 28B parameters) and supporting flexible input resolutions (224×224, 448×448, and 896×896), PaliGemma 2 is ideal for various tasks like detailed image captions… Continue reading PaliGemma 2 – New vision language models by Google

Quantum computing is rapidly evolving, with big challenges and bigger breakthroughs. Automating the complex calibration process of quantum systems is a game-changer, as seen with tools by Quantum Machines, Rigetti, and Nvidia, which use AI to streamline operations. Other tech giants like Google, Microsoft, IBM, and AWS are advancing quantum capabilities, from error correction to… Continue reading Nvidia, Rigetti, Quantum Machines Deliver AI-Powered Quantum Computing

Fireworks has launched beta Speech-to-Text APIs using advanced Whisper v3-large models, which can transcribe one hour of audio in just 4 seconds! Key features include real-time transcription, translation into multiple languages, and precise text-to-audio alignment with timestamps. It also enhances poor-quality audio with built-in tools. Fireworks’ APIs are fast, cost-effective (10x cheaper than competitors), and… Continue reading Fireworks Launches Fast Speech-to-Text API

Meta’s Llama 3.3 is a cutting-edge multilingual AI model optimized for text-based tasks like chat, coding, and reasoning. It supports 8 languages, uses a powerful transformer architecture, and has a long memory to handle large conversations. Trained on vast data, it excels in benchmarks, outperforming many AI models. Safety measures, such as human feedback and… Continue reading Meta’s new Llama 3.3 70B Instruct model

The Reserve Bank of India (RBI) is using cutting-edge AI technology called MuleHunter.AI to fight financial fraud. Mule accounts are used by scammers to move stolen money, often through people who unknowingly help, called “money mules.” MuleHunter.AI detects these accounts by analyzing unusual activity and transaction patterns in real-time. This allows banks to block such… Continue reading New AI tool MuleHunter.ai by RBI to help reduce digital frauds

Southeast Asia is embracing AI advancements with a focus on sovereign AI, where countries use local resources for AI development. NVIDIA’s CEO, Jensen Huang, met with leaders of Thailand and Vietnam to discuss initiatives supporting AI innovation. In Thailand, NVIDIA is partnering with universities and startups to boost AI for weather prediction, healthcare, and more.… Continue reading Thailand and Vietnam Embrace Sovereign AI to Drive Economic Growth

This paper introduces a method to create a better large-scale dataset for training large language models (LLMs) by combining techniques like classifier ensembling, synthetic data rephrasing, and reduced reliance on traditional filters. The approach results in a 6.3T token dataset with more unique, high-quality tokens compared to previous datasets like DCLM. This improves model performance… Continue reading Nemotron-CC: Transforming Common Crawl into a Refined Long-Horizon Pretraining Dataset