No Permission

I recently posted an exploration of the elements of the infamous Milo demo from E3. That post primarily focused on what took place in the demo, and if all was as it was represented to be.

But let’s assume for a moment everything in the Milo demo was “real”, that it all was what it seemed to be. What’s the purpose of Milo? What’s the goal? What’s Milo meant to be?

Milo would seem to be a digital companion, a software entity that’s intended to interact, learn, and grow with the user. The software has the avatar of a human boy, which helps considerably in building the bond with the user. And the software itself seems to learn about its users and learns to interpret their moods and emotions.

Milo also seems designed to trigger nuturing responses in the user. Molyneux, Milo’s designer, has said that Milo simply won’t respond to abuse. If Milo only responds to positive behavior, will that encourage users to be positive and nuturing with him? Will it change the users? Bits such as the exchange about completing homework seem to indicate that’s the case. And don’t forget the whole drawing a fish bit is also related to helping Milo with his homework. You can easily see how this trick could be used to get kids to do their own projects, or to motivate adults with theirs (Milo: “I just can’t seem to get this pivot table right in this spreadsheet. How will I ever calculate EBITDA?”).

But beyond simple motivational tools, what can move software like Milo to be more than the latest digital pet and to something more like an actual companion? It’s not hard to imagine a next step of other special-purpose Milos, designed properly with enough affective computing tricks to address issues like social anxiety or mood issues.

The ultimate Milo, though, would be adaptive, reacting to all sorts of general-purpose needs of the users. A true digital companion would learn and grow with its users, reacting in many ways as a real-life friend would. The general outlines of such a design don’t seem too complicated. The devil, as always, will be in the details.

There are issues with this, of course. There’s potential of misuse, both by the users and by the developers. Imagine a cult leader companion, for example, preying on weaknesses and convincing users to send money somewhere (don’t think it won’t happen!). Or a companion that encouraged anti-social behavior, or one that demanded all the users time. Or even just a badly designed companion with harmful bugs.

There’s also the issue of people interacting more and more with software instead of actual people. Is that a good thing for society? Personally, I think that aspect is a bit of a moot point, since it seems inevitable. It seems to me the correct approach is to shape the interactions so that they are socially beneficial.

This can all seem pretty pie in the sky, but when you consider what exists today and what can be done with some basic tricks that obviate the need for genuine AI, it’s probably only a few years until these start to become practical questions. I’m somewhat surprised we aren’t already seeing some more sophisticated Facebook and Twitter bots playing the affective/emotional games.

Of course, it’s things like Milo’s eye contact and his vocal cues that really make the difference. Humans are simple animals on a basic emotional level, and once these basic computing interface tricks become widespread, we’re in for a whole new future.