Your daily source for AI news, tools reviews, and practical guides to artificial intelligence
Following OpenAI’s recent disproof of the Erdős unit-distance conjecture, Anthropic’s AI, Claude Mythos, reportedly provided a simple and elegant proof, highlighting rapid advancements in AI-assisted mathematical problem-solving.
Research from Peking University highlights a pervasive issue in leading AI models like GPT and Gemini, where correct answers are often backed by inaccurate or unrelated citations, posing risks especially in critical fields such as law and medicine.
Ineffable Intelligence, a UK-based AI startup founded by former DeepMind researcher David Silver, has raised $1.1 billion in funding, aiming to create autonomous AI systems that learn independently from human-generated data.
Yann LeCun, former Meta AI chief scientist, has launched AMI Labs with $1 billion funding to develop modular, task-specific AI systems, challenging the dominance of large language models.
Jerry Tworek, a former researcher at OpenAI, has founded Core Automation, an AI lab focused on pushing the boundaries of current AI architectures through advanced automation and innovative learning methods.
A recent study using ProactiveBench shows that most multimodal AI language models tend to guess answers rather than ask users for missing visual information, though reinforcement learning may offer improvements.
Alibaba’s Qwen research group has developed a novel algorithm that enables AI models to think more deeply by assigning rewards to each reasoning step proportionally, effectively doubling the length of AI thought processes.
Research from Nvidia, UC Berkeley, and Stanford highlights the limitations of AI models in robot control absent human-engineered building blocks, while introducing methods that enhance AI performance in robotics.
Researchers at Sapienza University of Rome have devised a novel, training-free approach to identify hallucinations in large language models by measuring subtle computational traces, promising improved detection and understanding of AI errors.
OpenAI’s latest GPT-5.4 model introduces a new metric called ‘CoT controllability’ to evaluate AI’s ability to regulate its own reasoning. A recent study reveals that AI models largely struggle with this task, a finding OpenAI considers beneficial for AI safety.