I have set myself on fire in many different ways. This is one of the safest.
And about Google Duplex... I have heard tech people and journalists talking about this for a year, and almost always people's most vocal reaction is to be offended that the system is trying to imitate a person with all of those "ahhs" & "ummms" (vocal disfluencies). People are literally going off on rants about how there need to be laws to protect us from being deceived by an AI pretending to be a human.
I think Chemda is the first person I've heard get it right. The stammers are just to make people not feel so stupid about misunderstandings, and to be more efficient at getting it's job done. The examples they show off don't have the system making mistakes - just situations where a person misunderstands something and the voice has to try again, but a lot of the time, the AI will not understand the person, and will not know if the person misunderstood something, or changed the subject, or is having a stroke, or whatever, and it just has to try again to get its message across in a way that will result in the person saying something that it can parse and understand. It would not be helpful for the voice to say "No. I meant THIS, not THAT." Just like people, it has to play along so that no one feels like the asshole.
And pretending to be human is a kindness. If a system like this called people and first told them it was an automated system, and then communicated in simple, direct phrases, and corrected mistakes, and had all of the information in the world available to it, I guarantee that would trigger the Frankenstein effect more than anything else, because people would be confronted by their own inadequacies.
Remember, condescension isn't a problem if you're too stupid to know it's happening. Ultimately, this is the best we can hope for as a species, that our AI robot overlords will treat us as sweet, retarded children.
|