The Full Duplex

The next big leap in AI user interface will be the AI that initiates contact, that calls you. The Google Duplex may well be the breakthrough on this. This is a live feature of Pixel phones now, which I don’t own so I haven’t tried this feature. But looking forward to it, sounds very cool, a bot that actually could make a reservation for me.

If you were not aware:

Today we announce Google Duplex, a new technology for conducting natural conversations to carry out “real world” tasks over the phone. The technology is directed towards completing specific tasks, such as scheduling certain types of appointments. For such tasks, the system makes the conversational experience as natural as possible, allowing people to speak normally, like they would to another person, without having to adapt to a machine.

It’s been a year since this announcement and we should assume another year or two before this type of thing is offered by all the major tech providers, and works really really well. Scary well.

So, if I am to #TrustNoRobot, how should I feel about Duplex and its ilk?

I’m ok with this. With one big reservation.

First, I’m ok because this is exactly what AI should be doing. Focusing on how it can assist the individual. Not on what ads to show us. But how it can help with the hassles of the world.

AI assistants do not feel or care about us, nor should they feign doing so. Caring is what people do. Serving is what tools do.

I think the Google Home Pretty Please feature is a travesty that trains children to respect these algorithms as Beings. It should be killed with fire. It will bring up a generation deeply confused about where to take direction from.

In the same manner, the reservation I have with Duplex is the attempt to simulate human vocal patterns. This is fraud, and will result in nothing but hurt feelings for real humans.

Think about cars. They sort of have a biological shape often, something organic, and some are almost anthropomorphic in having “hips” around their tires. Certainly the headlights look like eyes, etc.

And people bond with their cars. Even love them. But they don’t try to marry their cars.

Now imagine if cars evolved to look exactly like their owners. But instead of getting in them these robots carried you on their backs.

Don’t you think this would cause a lot of confusion? That maybe this would be a weird and horrible world to live in? That we should not do that as a goal of civilization? That at best this would be a show on Hulu that you don’t want to watch?

That’s how I feel about anthropomorphic AI. Let it be like a human, but always clearly difference in it’s user interface. Or let’s get creative and not make it human at all. But always: Signal to the user that this is an agent of another human, not only in clear words but also in presentation.

The UX folks at Google seem to think the goal is to not have to adapt to a machine. Not realizing that that our adaptation is what defines something as a tool or a person. Otherwise you create potential moral confusion.

And always represent only one other human Singleton. No AI bosses. No AI ever working on “behalf” of others who hide their identities.

With proper legislation, AI can be made as safe as highways. Which are still really unsafe, but beat dirt trails in dark forests hands down.

Because AI is a road.

Painting by Simon Stålenhag

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.