Let me preface this by first saying that I absolutely can't wait to have my own personal home automation, AI assistant, etc. on prem without the cloud:
I think that as far as the nascence of these features goes, the cloud model will beat the on-prem features any day of the week for several reasons. Lack of configuration to set up, ease of use from anywhere without network configuration, etc. are table stakes. But the biggest at this point is the sheer amount of training and A/B testing data you can ingest to determine what is useful for your end users.
The velocity of cloud-based products is nothing short of amazing and I doubt that on-prem will compete with the feature set and ease of use of always connected solutions until there are feature-complete, mature cloud versions to then bring in.
As we just learned with Yahoo, though, once the ML models have been trained, they can be disseminated and used without the need for "cloud-scale" data or compute resources.
And, for better or worse, Dragon's text-to-speech is pretty damned good after a rather minimal amount of training.
I don't think there's anything stopping voice and intent recognition from coming back to our personal machines other than the ability to keep making money from having it come up to the cloud.
I think that as far as the nascence of these features goes, the cloud model will beat the on-prem features any day of the week for several reasons. Lack of configuration to set up, ease of use from anywhere without network configuration, etc. are table stakes. But the biggest at this point is the sheer amount of training and A/B testing data you can ingest to determine what is useful for your end users.
The velocity of cloud-based products is nothing short of amazing and I doubt that on-prem will compete with the feature set and ease of use of always connected solutions until there are feature-complete, mature cloud versions to then bring in.