Voice recognition software (VRS) for medical transcription, for instance, has to make a number of arbitrary interpretations that artificial intelligence (AI) may not yet be fully capable to tackle perfectly.