It may be a mistake to think of this as robotics. Consider it more in terms of form-factor leading to UI interaction styles: - mac for the desk - ipad for th

It may be a mistake to think of this as robotics. Consider it more in terms of form-factor leading to UI interaction styles: - mac for the desk - ipad for the couch - iphone for the pocket and hand - Vision headset when you are inside the computer So what's the computer you talk to from afar, or point to, to have it do things? Where's the screen that will follow you? To me the sea change would be moving to voice as the primary mode of interaction. Keyboards are just awkward and hard. And voice now is used only with screen-centric UI's, as a keyboard replacement. The new thing could be life beyond the thumb. For tasking a robot, the UI is more like a dialog, of setting objectives and constraints and getting feedback on progress. The key is the level of familiarity and trust: what might start as awkward and verbose with lots of confirmations, becomes laconic with experience of repeated success. That's the stickiness. Transaction cost economics describes how repeated auctions accrue advantages to incumbents. Everyone's task dialogs will build differently, and once you've taught your assistant to be functional, you won't want to re-teach another; and also they'll likely better learn how you want the next task done. That's the value of idiosyncratic on-device intelligence, intelligence that's not just fodder for consumer targeting or model building. People will pay extra not only for quality, but also for privacy, especially in the intimacy of tasks.