For years most people have been able to order food online from both takeaways and supermarkets. I've also been able to book doctors appointments online in the UK online for years. In short I don't need this special technology, to me it's just a gimmick apart from for some disabled people and typing in the traditional way is still much more accurate for me, but voice recognition has been around for years too if I really wanted to use it.
In fact I tried voice recognition for the first time in the 1980s and speech synthesis in the early 1980s which was surprisingly quite good even then. Yes the technology has improved, but speech recognition has still got a long way to go before it can understand people as well as a person can without constantly making mistakes, a limited vocabulary can greatly help, E.g. a limited set of specific commands, but if it needs to recognise most words in the dictionary, many of which that sound similar, it will really struggle. If you were for instance setting an alarm on the Amazon device, the system would then only expect a limited number of words to differentiate against after it knew you were setting an alarm, so this will be fairly successful. If you are taking a dictation however you need a certain amount of common sense and the artificial intelligence equivalent is far from adequate at the time of writing, for instance a computer will often struggle to know which spelling of a word to use in the current context, E.g. should it write "deer" or "dear", "their" or "there", "B. Wright" or "be white" or even "bee write" Etc., many AIs will attempt to choose the more likely word spelling based on the words before it, but it will still often get it wrong because an AI won't have a full understanding of what is being written. It will also often struggle a lot more if you don't talk carefully to it at the same volume without changing expression and background noise can be an issue too (a headset can help).
From 1986:
So you can see the technology is far from new and speech synthesis was around a lot earlier still, obviously we didn't have the Internet in those days however, although there was other much less used information networks even in the 1980s. This speech recognition system even attempts to differentiate between words that sound the same with different meanings and spellings by using a limited AI. They however make the system appear so much better than it is, in reality it will constantly make mistakes and for a reasonable typist, using the keyboard will be much faster as it still is even today.