Patients at GP practices in Rotherham are hanging up on an AI receptionist called Emma after it fails to understand their local accents. Health watchdog Healthwatch Rotherham has flagged the issue despite the AI firm claiming support for 17 languages.
Several medical practices in South Yorkshire have deployed the AI receptionist to handle patient calls, but the system struggles with broad regional accents common to the area.
Healthwatch Rotherham, which monitors local health and social care services, reported that patients are becoming frustrated when Emma cannot process their speech patterns and accent variations—described as local "twangs" by the watchdog.
The AI firm behind Emma states the system supports 17 languages, yet appears unable to reliably handle accent diversity within a single English-speaking region. This limitation highlights a broader challenge in AI voice recognition technology: systems trained on standardized speech patterns often falter with regional dialects and accent variations.
The issue raises questions about the readiness of AI-powered customer service tools for real-world deployment in diverse communities. GP practices implementing such systems without adequate testing for local speech patterns risk degrading patient experience and accessibility.
Healthwatch Rotherham's findings suggest the need for more comprehensive testing before rolling out voice AI in healthcare settings.
OpenAI's recent solutions to longstanding mathematical problems have triggered significant debate within the mathematics community, with leading mathematicians grappling with the implications of AI's capabilities.
Chinese AI models Kimi K3 and GLM-5.3 have closed the gap with leading Western alternatives, forcing a reassessment of where competitive advantage actually lies in artificial intelligence.
Generalist AI has released GEN-1.5, an artificial intelligence model capable of teaching robots new tasks after observing just one demonstration. The breakthrough addresses a major limitation in robotics: the need for extensive training data.