Comment Re:Purely out of curiosity (Score 5, Informative) 692
(Reply written before most other posts, was distracted by work, going to post anyhow even with some now redundant info. Hope it helps.)
Android's voice recognition is mostly a search input box, driven by voice instead of text. It's pretty clever how Google built the system, they used voice input from the old GOOG411 number to help adapt it to different languages and accents. For the most part though, it will parse what you say and do the equivalent of "I'm Feeling Lucky" on google.com.
It also does dictation for typing in notes, or other apps. Basically anywhere the keyboard will appear, voice can be used as a dictation input.
Siri is a step beyond what Google offers, due to the conversational style of input vs just basic voice commands/dictation. You can say "Joanne Moore is my mother" to Siri once. Later, saying "Text Mom that I'll be late for dinner", and Siri remembers mom = Joanne Moore, or whoever. This just scratches the surface, the other power of Siri is the capability to understand questions like "Do I need a raincoat today?". It turns that into a search of the weather at the current location, scanning the days forecast for the possibility of rain. A followup of "what about Saturday?" would cause Siri to recognize this is a followup request, and it would link it to the previous weather query. The logic is in the Siri system, not in a search engine being queried. Minor detail, and either approach can work.
Google can improve their services on Android by improving what Google.com does, and this benefitting web users as well. For Apple, they have to decide what services to tie into. Many queries in Siri are farmed out to Wolfram Alpha. http://en.wikipedia.org/wiki/Siri_(software) has more info on other services it integrates with to try and answer questions. If none of those work, it defaults to running a web search similar to Android.