AI is in a constant loop of some silly scandals. Everyone talks about this tech as of the future, which has been around the corner for years now. Indeed, AI has given a lot to us and businesses already; therefore, no wonder there is so much trust and hope in this technology.
However, the drawbacks of AI always turn out to be a bit scandalous because if something goes wrong, it is either awfully bad or hilariously wrong: two occasions that take no time to dominate the media.
In both instances, the drawbacks come from the fact that AI has no consciousness; it isn't good and reading the context. This also makes it quite shabby to do automated transcriptions or translations, and we will explain why.
When humans listen to a speech, they reflect on the words in sentences, paragraphs, and context. Even poorly articulated someone is easy to understand once we consciously put all the clear and murky pieces of the puzzle into a single picture. Therefore, context matters.
But can AI assume the correct meaning of words or sentences, then it just tries to identify a word that has been said in an isolated manner? Transcription and translation are so much more beyond words. It is all about the meaning.
Especially if the recipients of the transcribed content are foreigners who cannot catch that something went wrong in the text. Or people with hearing impairment may see the captions made of the transcription and become quite puzzled about the intended message.
Devil is in the details!
2. Selecting the right word
Now try to choose the right word when you have no idea of a context:
- wail, wale, whale
- knave, nave
- cite, sight, site
- peace, piece
And there are almost 500 such words, which are called homophones. You don't want to mess these up. It is much easier by understanding the context to catch the right word instead of leaving it to a random chance that AI will pick and choose the correct spelling.
At GoTranscript, we employ a professional team of native transcribers and editors. GoTranscript delivers top quality for general and specialized transcriptions. Check it out.
Curious to learn more about what may go wrong with AI transcriptions? See Part 2.