I’ve always been extremely skeptical about speech recognition. Like machine translation, it’s always been one of those technologies that for some reason is the “coming great thing” for a certain group of tech pundits, while not getting any closer (or even steadily receding) when you look at actual implementations. Here’s a particularly hilarious example. More here.

Understanding spoken language is a problem even for people. I listened carefully and, frankly, didn’t quite catch several sentences spoken by the operator in that first video. Even though he tries to avoid running words together as he would probably do when speaking to another person, he’s prone to leaving vowels out (“c’rrect”, “d’let'”, “cap’t’l”). Personally I’m able to misunderstand people in several languages, which actually adds to the problem – before decoding the words, I have to identify the language. Great fun on trips.