We’re still a ways away from perfect voice interaction but within 3–5 years, it’s likely that all voice assistants:
- Will work at less than 1% WER
- Will have far field modes to eliminate a lot of local requirements
- Will have sub second response times
- Will work across many domains of knowledge
- Will have APIs for easy integration in third party hardware
Basically, the time for a new entrant to go from scratch to Alexa-like AI assistant will be quick — maybe a few months (or maybe less for brand-able versions).
Consumers will likely have to pick their favourite persona for voice interaction and they might have to weigh the assistant that brings the most features and integrations for them specifically. New features and games might be necessary to lure users from one AI assistant to another (Bixby’s points is an interesting first stab at this).
What is likely is that there will be a long tail of AI assistants that will work much better than Google Assistant and Alexa work today.