Google's artificial intelligence-powered medical chatbot has achieved a passing grade on a tough US medical licensing exam, but it's answers still fall short of those from human doctors, a peer-reviewed study said on Wednesday.