Artificial intelligence is demonstrating notable impact in healthcare and the legal sector. Ambience Healthcare has introduced an OpenAI-powered ICD-10 medical coding model that outperforms physicians by 27% in coding accuracy. The model was benchmarked against 18 board-certified doctors and is already deployed at more than 40 organizations, including UCSF Health and Cleveland Clinic. Ambience used OpenAI tools and reinforcement fine-tuning to develop the system, which supports ICD-10 and CPT codes. The company has raised over $100 million, with speculation of a valuation above $1 billion, and plans to roll out the new model to customers before fall. In emotional intelligence, Swiss researchers have found that AI systems now surpass humans in understanding and managing emotions, as well as in interpreting emotional nuance in text. AI models are able to both solve and create emotional intelligence tests. In the legal field, a database compiled by Damien Charlotin has tracked 120 court cases worldwide since June 2023 where AI-generated 'hallucinations'—fabricated or incorrect legal citations—were submitted in court documents, including more than 20 cases in the past month. Notable incidents include the Mata v. Avianca case and recent Ontario court cases, resulting in financial penalties, formal warnings, and public reprimand for lawyers. Law firms are responding by training partners to supervise AI use and integrating digital literacy into legal education. Research indicates that in-house lawyers are increasingly comfortable with generative AI tools for contract review and compliance monitoring. As AI adoption grows, legal professionals are encouraged to focus on critical thinking and empathy, while maintaining oversight of AI-generated content.
Large language models are proficient in solving and creating emotional intelligence tests https://t.co/57N2q97hCC
New website. Let's get every radiologist an AI assistant. https://t.co/UQFNKkpWjC
LLMs generating radiology reports struggle with clinical accuracy due to knowledge limitations and hallucinations. This paper introduces RADAR, a framework integrating both the model's internal knowledge and external information to enhance report accuracy. Methods 🔧: → RADAR https://t.co/giwDGx7W0a