> Amazon is betting way too heavily on voice as the optimal way to interact with hardware. I know that it's cool because it hasn't ever been done well before, but in practice the utility just isn't there for the vast majority of use cases.

Beyond the issues with dialects and accents - those will eventually be solved with better learning/adaptation algorithms - I've always felt voice interfaces very... limited. Perhaps it will enable more 'social' experiences, but 3-5 people can be in a room and use their devices (laptop/tablet/etc) more or less as they do now. If/when devices with 'voice interactions' are the norm... how will I get any privacy from those around me? And simultaneous voices will present some ongoing problems (until, maybe, personal voice recognition, vs just 'voice', is achieved?)