All AI news
Browse, filter, and search every article in the archive. The homepage shows the last 24 hours; everything older lives here.
A New Framework for Evaluating Voice Agents (EVA)
Hugging Face has introduced a new framework called EVA for evaluating voice agents, aiming to provide a standardized method for assessing their performance. This framework is designed to enhance the development and deployment of voice AI technologies by offering clear metrics and evaluation criteria.
Measuring progress toward AGI: A cognitive framework
Google DeepMind has introduced a cognitive framework aimed at measuring progress towards artificial general intelligence (AGI). This framework provides a structured approach to evaluate AI systems' capabilities and their alignment with human cognitive processes.
ImportAI 449: LLMs training other LLMs; 72B distributed training run; computer vision is harder than generative text
The article discusses recent advancements in large language models (LLMs), including the training of one LLM by another and a significant 72 billion parameter distributed training run. It also highlights the challenges faced in computer vision compared to generative text models.
From games to biology and beyond: 10 years of AlphaGo’s impact
The article reflects on the decade-long impact of AlphaGo, highlighting its influence across various fields, including games and biology. It discusses how the advancements in AI, initiated by AlphaGo, have led to significant developments in problem-solving and decision-making processes in diverse domains.
Import AI 448: AI R&D; Bytedance's CUDA-writing agent; on-device satellite AI
The article discusses recent advancements in AI research and development, highlighting Bytedance's new CUDA-writing agent and the emergence of on-device satellite AI technologies. These innovations reflect the ongoing integration of AI into various sectors, enhancing capabilities and efficiency.
Import AI 445: Timing superintelligence; AIs solve frontier math proofs; a new ML research benchmark
The article discusses advancements in AI, including the timing of superintelligence, AIs successfully solving complex mathematical proofs, and the introduction of a new machine learning research benchmark. These developments highlight the ongoing progress in the field of artificial intelligence and its implications for future research and applications.
Project Genie: Experimenting with infinite, interactive worlds
Google DeepMind has launched Project Genie, an initiative aimed at creating infinite, interactive worlds through advanced AI techniques. This project explores the potential of AI to generate dynamic environments that can adapt and evolve based on user interactions, pushing the boundaries of immersive experiences.
D4RT: Teaching AI to see the world in four dimensions
Google DeepMind has introduced D4RT, a new AI framework designed to enhance the perception of AI systems by enabling them to understand and interpret the world in four dimensions. This advancement aims to improve AI's ability to process complex spatial and temporal information, potentially leading to more sophisticated applications in various fields.
Import AI 439: AI kernels; decentralized training; and universal representations
The article discusses advancements in AI kernels and decentralized training methods, highlighting their potential to enhance model efficiency and performance. It also explores the concept of universal representations in AI, which could lead to more versatile and adaptable models across various applications.
Google's year in review: 8 areas with research breakthroughs in 2025
Google DeepMind reflects on significant research breakthroughs achieved in 2025, highlighting advancements across various domains of artificial intelligence. The report emphasizes the impact of these innovations on the future of AI technology and its applications.
Gemma Scope 2: helping the AI safety community deepen understanding of complex language model behavior
Gemma Scope 2 is a new tool developed by Google DeepMind aimed at enhancing the AI safety community's understanding of the behaviors exhibited by complex language models. It provides insights and analysis to help researchers and practitioners better navigate the challenges associated with AI safety.
Import AI 435: 100k training runs; AI systems absorb human power; intelligence per watt
The article discusses the latest advancements in AI training, highlighting the completion of 100,000 training runs and the increasing efficiency of AI systems in harnessing human power. It also touches on the concept of intelligence per watt, emphasizing the importance of energy efficiency in AI development.
It’s the Humidity: How International Researchers in Poland, Deep Learning and NVIDIA GPUs Could Change the Forecast
International researchers in Poland are leveraging deep learning techniques and NVIDIA GPUs to enhance weather forecasting, particularly focusing on the impact of humidity. This innovative approach aims to improve the accuracy of predictions, potentially transforming meteorological practices.
NVIDIA Research Shapes Physical AI
NVIDIA Research is advancing the field of physical AI by developing technologies that enable machines to interact with the physical world more effectively. This research focuses on improving the understanding of physical dynamics and enhancing the capabilities of AI systems in real-world applications.
A Gaming GPU Helps Crack the Code on a Thousand-Year Cultural Conversation
NVIDIA's latest gaming GPU has been instrumental in analyzing and interpreting a complex cultural dialogue that spans a millennium. This technological advancement showcases the potential of AI in understanding historical contexts and cultural narratives through enhanced computational capabilities.
What Are Foundation Models?
The article from NVIDIA AI explains foundation models, which are large-scale AI models trained on diverse datasets and capable of performing various tasks. It discusses their significance in the AI landscape, including their applications and the advancements they enable in machine learning and natural language processing.
AI-Designed Proteins Take on Deadly Snake Venom
Researchers have utilized AI to design proteins that can neutralize the effects of deadly snake venom, showcasing the potential of AI in biomedicine. This innovative approach could lead to new treatments for snake bites and enhance our understanding of protein interactions. The study highlights the intersection of AI technology and healthcare advancements.
What Is Retrieval-Augmented Generation, aka RAG?
Retrieval-Augmented Generation (RAG) is a technique that combines retrieval and generation to enhance the performance of AI models in producing relevant and accurate responses. By leveraging external information sources, RAG improves the contextual understanding of generated content, making it particularly useful for applications requiring up-to-date information. This approach represents a significant advancement in the capabilities of AI systems in handling complex queries.
AI Maps Titan’s Methane Clouds in Record Time
NVIDIA's AI technology has successfully mapped the methane clouds on Titan, Saturn's largest moon, in a record time, showcasing the capabilities of AI in planetary exploration. This achievement highlights the potential for AI to enhance our understanding of extraterrestrial environments.