I saw this article on Mashable and thought it was worth comment:
Speech recognition software has been around for a long time. I have been using Dragon Naturally Speaking since the late 90s. The software is pretty awesome and during my undergraduate years was able to speak the first draft for most of my papers (I would have to go in to correct the mistakes after the first version). I also found this software extremely helpful when I had to transcribe videos in both corporate and educational settings. To transcribe, I would listen to the speech and talk as they were talking. It was a very quick way to transcribe.
Currently I use speech recognition for two things on a regular basis: talking to my car and my phone. I have the Ford Sync system on my car. So whenever I want to call someone I tell my car to call them. Then I speak to the person through my speakers. It is very useful. Then on my phone I speak to my Galaxy Nexus for many of my texts.
How accurate is speech recognition? Honestly, I notice no improvements since the late 90s when I started using it. It is somewhere around 90% (guess not a true stat). If there is noise it is even worse. I have also noticed that it just does not like the way some people talk. I talk very clear and so it works for me. I have also noticed that accents do not seem to affect it, it’s more about the clarity of the persons voice.
So will speech recognition get more popular? Well it has with SIRI. However, are people still using SIRI? Many of my friends found it annoying after the honeymoon period. That does not mean it is not useful. I do see it becoming more popular as people adapt to the idea of speaking to their computers or phones. However, it is still not perfect and I am not sure it will ever be as we just do not have the technology for perfect AI at this time, so until then the mistakes may turn off users.