Explore 6 articles on the latest research and developments in AI
A new study reveals that simple rule-based AI methods outperform advanced language models in extracting patient details from Polish medical texts, highlighting critical trade-offs for global healthcare.
New research shows that AI-generated code often contains hidden flaws that make it hard to maintain, but simple changes to prompts and models can significantly reduce these problems.
A new method uses the mathematical study of shapes to clean up manipulated text images, allowing OCR systems to read accurately even when attackers try to fool them with invisible changes.
A new hypothesis-dependent noise condition allows AI classifiers to achieve the same strong performance bounds as before, but under less restrictive data requirements, broadening real-world applicability.
A new taxonomy helps organizations track AI dependencies in critical fields like healthcare and law, preventing errors from opaque data and models.
New technologies are making banks active partners in detecting money laundering, cutting false alarms by 30% and reducing manual work by over 50%, but raise governance challenges.
