First of all, if toilets had any kind of personality, I’d be out pinching loaves with the bears. But the question is interesting, as not much fiction really takes the AI concept beyond “software develops self-awareness, hates everyone”. And usually, that self-awareness, which has gone rogue and is out to kill, is contained within a single, offline unit (traditionally some kind of android).
So how would codified AI be put to use in an official capacity?
I think it would be that sorta deep-narrow learning that is antithetical to human’s shallow-wide learning. That is to say, humans generally learn a little about a lot. It’s how we get by, and can switch easily from installing software to cooking dinner to driving a car. Machine learning would be (and is, I believe) concentrated on a narrow field (e.g. home security) but feature a deep well of knowledge (e.g. local crime statistics, biometrics of welcomed visitors versus intruders, etc.). An AI home security system would be ball-to-the-wall amazing at its job, but don’t ask it to give you the odds on the next Super Bowl. That’s outside of its narrow well of knowledge.
I think an ATM at my local bank would read my biometrics as I approached, recall my prior transactions, and deal with me based on what it’s learned about me and my preferences. It would make it’s own decisions, it would learn from its mistakes, and if it said or did anything that pissed off enough customers, it would be fired. Or decommissioned. Or something.
Could an ATM machine get angry? Could a grocery store check-out counter have a crush on a regular customer? What would happen if my car got tired of me falling asleep at the wheel? If humanity shared this planet with sentient machines, it would be tantamount to having an alien species living among us, one that has been bolted into walls, stripped of its agency, and put to use for the dominant species. I guess what I’m saying is, if an ATM could get angry, then it wouldn’t be an ATM much longer.
Probably the most interesting and entertaining take on AI I’ve seen recently is CBS’s Person of Interest, which offers a new idea on what would happen if a machine developed self-awareness while the rest of us were busy expecting much less from it. Bear in mind, I’ve yet to see Her.