Tag: artifical intelligence

I use my “AI” per­son­al assis­tants inter­mit­tent­ly, most­ly because they aren’t very good assis­tants; yet. For these things to actu­al­ly get use­ful, they need to be able to do more than build lists, order things, send reminders, and pro­vide wayfind­ing.

Use Case: Adding a stop to my wayfind­ing

  • Exist­ing Func­tion­al­i­ty: I ask: “Dunkin Donuts along my route”. I’m told “Find­ing Dunkin Donuts along your route along your route.” and pre­sent­ed with a list of x num­ber of loca­tions that I have to scroll through, deter­mine which is the best option, and add to the wayfind­ing while I’m dri­ving. 
  • Bet­ter Func­tion­al­i­ty: I ask “Dunkin Donuts along my route.” and the assis­tant choos­es from the options avail­able and auto­mat­i­cal­ly adds the one with least impact on my trav­el time to the wayfind­ing.
  • Best Func­tion­al­i­ty: I say “Order my usu­al from the near­est Dunkin Donuts on my route” and the assis­tant: finds that loca­tion, places an on-the-go order using infor­ma­tion it already has, and sched­ules the pick-up based on trav­el time to that loca­tion.

Use Case: Reminders

  • Exist­ing Func­tion­al­i­ty: I say: “Remind me to get up at 1am on Sun­day for the Ori­on­id mete­or show­er”. I get a noti­fi­ca­tion at 1am on Sun­day. I sleep through it.
  • Bet­ter Func­tion­al­i­ty: I say: “Remind me to get up at 1am on Sun­day for the Ori­on­id mete­or show­er”. An alarm is set instead of a push noti­fi­ca­tion, because the assis­tant knows I’m usu­al­ly asleep at this time.
  • Best Func­tion­al­i­ty: I say: “Find me the best place and time to watch the Ori­on­id mete­or show­er on Sun­day.” The assis­tant knows my loca­tion, checks the Dark Sky data­base for near­by loca­tions with low­er light pol­lu­tion, cal­cu­lates trav­el time to that loca­tion for best view­ing, and sets an alarm to wake me up so I can get there on time.

Oth­er stuff that would be cool:

  • “Pay my gas, elec­tric, inter­net, and phone bills on Fri­day.” This finds the bills, and sched­ules pay­ment with my bank.
  • “Send $50 worth of flow­ers to my mom the Sat­ur­day before every Moth­er’s Day” Bonus points if it knows what kind of flow­ers to send.
  • “Add ‘The His­to­ry of Phi­los­o­phy’ pod­cast to my morn­ing dri­ve Spo­ti­fy playlist”
  • “Sched­ule 3 days of weight train­ing and 3 days of car­dio, and adapt the sched­ule based on missed work­outs.” This isn’t just putting some­thing on the cal­en­dar, but is active­ly pro­vid­ing the work­out details as well.
  • “I want Ital­ian for din­ner twice this week.” This would find a recipe or two based on how much time it knows I have to pre­pare din­ner, and auto­mat­i­cal­ly add the ingre­di­ents to my list. Bonus points if it can ana­lyze past meal plans & pre­dict what ingre­di­ents I already have on hand.

These AI assis­tants are way less capa­ble than my child of under­stand­ing & inter­pret­ing con­ver­sa­tion­al speech and using con­text & ini­tia­tive to pro­vide a qual­i­ty return. If you don’t ask the ques­tion using the exact phras­ing they know, you’re wast­ing their time. Assis­tants waste my time if I have to learn their lan­guage in order to get half-func­tion­al results from them.

They may get smart one day, but the walled-gar­den par­a­digm is going to make it a pain in the ass. My AI of choice should work with all of my oth­er appli­ca­tions of choice, but I have a feel­ing it will be less by choice & more by neces­si­ty in the long run. I can’t see Google’s AI inte­grat­ing with Apple to down­load some­thing in iTunes. Or Alexa allow­ing you to order some­thing from an Ama­zon com­peti­tor.