The thing is that 'Likely' is very far away from 'always'. There is no guarantee the mistake will be easy to spot. For entertainment purposes AI transcription

The thing is that 'Likely' is very far away from 'always'. There is no guarantee the mistake will be easy to spot. For entertainment purposes AI transcription is awesome. For serious business applications the ability to recognize mistakes will continue to be a field to which serious attention is given. It would be interesting to see AI processes double check itself, and also run a logic check on whether the transcription makes sense. So that it can report sections flagged as incongruous or of dubious reliability.