The main problem now is not speech recognition. It's a kind of uncanny valley effect. You can speak to these assistants, but the language is still restricted.

The main problem now is not speech recognition. It's a kind of uncanny valley effect. You can speak to these assistants, but the language is still restricted. They show little to no common sense. It's a lot of party tricks bundled together. You can't interrupt them and it's hard to correct them. On the speech recognition side, an issue I've found (although is a rather niche one) is triggered because I'm bilingual (I'm fluent in spanish and english). Speech recognition only works well on a single language. I have Alexa set to speak english, for example when I'm searching for a song with a spanish title, I have to try to fudge the name into a fake english pronunciation for it to produce the right phonemes that will match the song title, rather than just say the name properly. Also, if it misses the match, there's no easy way to stop it and say "No, not that one", and be presented with a list of similar matches.