"the technology is improving"
Undoubtedly, but it still fails with way too much regularity on anything that is not the simplest of soundbytes.
One day, we will undoubtedly have speech recognition capability that can actually recognize what a person is saying with a 95% or better success factor, but expert systems have to be fine-tuned and it has already taken years to get them where they are. I'm not holding my breath for this.