Dumb Design & Dumb AI
"Smart" design and technologies do not imply intelligence; it implies perceived complexity.
Merely responding to a limited set of inputs with pre-determined actions and/or communicating on a protocol—like Bluetooth or connecting to a home network—makes "smart" technology quite dumb as these features are more ubiquitous.
If the bar for 'smartness' is constantly changing due to contrasting features against other devices in the market, it's the wrong standard—revealing smart is marketing fluff.
Smart should have been replaced with convenient; "convenient devices" sounds much more like the time-saving, all-in-one, always-connected marketing terms with which they are associated. Regardless of the label, the goal of this technology is not to maximize function-appropriate behavior like playing go or recognizing cars and speed signs but to improve day-to-day human life.
The next domain of human-life improvement is inter-human collaboration. Within heterogeneous social systems, actors often self-regulate in order to minimize tensions or maintain power across different scales. For example, this might mean electing a government to enact laws or agreeing that it's your turn to take out the trash on Thursdays. Where there are unsymmetrical nodes in the network, there are unsymmetrical nodes or connections between them.
The increasing development of technology or AI as new agents or nodes within these social systems creates significant effects. An underperforming actor within a social system drives others to outperform and overachieve to maintain or better the status quo. Unsurprisingly, this applies to developing 'Dumb AI' to do the same.
Is the new "smart" technology spontaneous in output, either behavior or the human response to said output? Yes, it will be surprisingly dumb. It's likely to be technology interactions that are optimized to do enough to help us humans help ourselves.