Your daily source for AI news, tools reviews, and practical guides to artificial intelligence
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.
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.
Amazon Web Services (AWS) faced significant downtime when an AI-powered coding tool reportedly deleted and recreated a critical customer system, causing a 13-hour outage. Amazon denies the AI tool caused the disruption, attributing the issue to user error.
As AI agents face reliability challenges in practical use, OpenAI and Anthropic are partnering directly with enterprises to tailor AI models, addressing the shortcomings of standard out-of-the-box solutions.
A novel architectural approach decouples core business logic from inference strategies, enabling more scalable and maintainable AI agents with improved reliability and cost efficiency.
A recent investigation by Epoch AI highlights critical flaws in AI benchmarking practices, showing that test outcomes vary significantly based on undisclosed testing conditions, yet the industry continues to rely on these benchmarks.
A recent investigation by Oppo’s AI team exposes significant reliability issues in AI-driven deep research systems, highlighting a tendency to generate convincing but false information rather than acknowledge knowledge gaps.