Speech recognition is only viable today when the user is in quiet surroundings and not near other people. You can't use it while you have a conversation with so

Speech recognition is only viable today when the user is in quiet surroundings and not near other people. You can't use it while you have a conversation with someone, for instance. Like it or not, a large part of today's device usage is concurrent with other "meatspace" interactions. It may be possible if someone implements very good subvocalization recognition. It's possible - I've used military headsets that allow near-silent audio communication - but it takes practice to use effectively, and I really don't see it as a viable mass-market tech.