I am in no wise philosophically opposed to using such a solution. In my Day Jobbe life, I have more than a passing familiarity with Automated Speech Recognition (ASR) technology. It's wonderful stuff and can be quite powerful. I like the concept plenty. ASR can be liberating on a number of fronts, from the narrowly technical to the profoundly creative. There's only one small problem.
I don't talk like I write.
For a long time I've been of the opinion that if you stuck a professional writer's head into an fMRI machine (presumably whilst still bolted to the rest of the professional writer in question), you'd find that the speech center which lights up when composing fiction is distinct from the speech center used for ordinary, everyday communication. It's English as both a first language and a second language. In my case, my written fiction syntax and style are noticeably different from my spoken syntax and style. Sentence length and complexity, word choice, rhythms — I'm two different people.
The writer who's been in careful training since 1990 is a different speaker than the blabbermouth who might use Dragon. The stories each of me can and would tell are quite different.
So while I'll turn to Dragon if I can, once my hands give out if they do, I don't want to go there too soon. I'll lose something essential. I might gain something just as wonderful — I am open to the possibility — but right now I value what I have while I still have it.