I expect them to improve because there are obvious, straightforward, practical improvements they can make already. As an obvious example, the NLP is currently limited in terms of the variability of ways to add things to your shopping list. Adding more variants of that would be easy.
The huge amounts of data they're gathering about what people actually use this for is indeed a huge advantage. With that kind of data, they can build a centralised dictionary of common patterns for asking just about anything. Because of Siri's practical orientation, there is no need to build a "generic NLP engine" (which is what you're, quite rightly, suggesting is slow to improve). They merely need to improve the effectiveness of the NLP engine they already have, by making it understand more common patterns. That's achievable through expert-system-like functionality that's existed for 30 years.
In terms of the voice recognition, that has been progressing steadily for 30 years too, and I imagine Siri will benefit from improvements to the technology base, as others will.
There's a tension: the great advantage Siri has is a constrained domain (as you note); adding patterns, services and "ways to add things" etc expands this domain. (And adding new patterns, if they amount to grammar rules, can increase the domain dramatically.) It's a tradeoff between expressiveness and error.
My information is that the progress in voice recognition has been terribly slow, and in the last 10 years or so has been made mostly by limiting domains.
"In terms of the voice recognition, that has been progressing steadily for 30 years too, and I imagine Siri will benefit from improvements to the technology base, as others will"
Will it progress to a point where ambiguity is reduced to nil? The strength of the command line abstraction is its completely unambiguous interface - do what I say.
The do what I mean interface of Siri could be limited to actions with insignificant consequences - the human brain incorporates the best speech recognition available and still makes mistakes. Any speech AI performing significant actions would need to outperform the brain.
edit: That is not to suggest AI speech recognition outperforming the brain is not possible. There are probably metrics and methods for disambiguating common sources of confusion which could be performed in the blink of an eye rather than the minutes, hours or weeks later you find yourself thinking "Oh, THAT'S what he said!"
The do what I mean interface of Siri could be limited to actions with insignificant consequences - the human brain incorporates the best speech recognition available and still makes mistakes. Any speech AI performing significant actions would need to outperform the brain.
Or actions which can be undone.
I'd be very concerned if the US army decided to use Siri to control its nuclear missiles. Less so if someone uses Siri to change their thermostat or query their fridge contents.
"I'd be very concerned if the US army decided to use Siri to control its nuclear missiles. Less so if someone uses Siri to change their thermostat or query their fridge contents"
Indeed... A command line for non-critical parts of the world.
For launching missiles? Nothing less than a bash script! ;)
The huge amounts of data they're gathering about what people actually use this for is indeed a huge advantage. With that kind of data, they can build a centralised dictionary of common patterns for asking just about anything. Because of Siri's practical orientation, there is no need to build a "generic NLP engine" (which is what you're, quite rightly, suggesting is slow to improve). They merely need to improve the effectiveness of the NLP engine they already have, by making it understand more common patterns. That's achievable through expert-system-like functionality that's existed for 30 years.
In terms of the voice recognition, that has been progressing steadily for 30 years too, and I imagine Siri will benefit from improvements to the technology base, as others will.