King Ludd #22: Domesticated Intelligence

You won't want to micromanage your smarthome

Cabin Cabin

The idea of a smarthome that only does what the owner tells them to do is an oxymoron. Smart things are smart by virtue of the fact they make intelligent decisions, and do not need to be micromanaged. You want to be able to delegate certain tasks and trust that an acceptable solution will be reached.

The emerging mainstream vision of a smarthome boils down to little more than being able to control lights remotely, often via a non-optional cloud, which unsurprisingly is not proving too compelling, and not too Friendly. Often all that is being provided is a remote switch, without the potential for integrating the controls from multiple vendors into a single coherent interface. Yet once we arrive at that ability to interface so much together who is actually going to have the time and inclination to wire it all up and then program it to understand what they as a unique individual, with preferences that evolve, want it all to do in every conceivable situation? Those people have already built their own home automation systems.

From the perspective of the currently unautomated dwelling inhabitant end user, the ideal smarthome product arrives in a box, you perform some trivial installation procedure, such as plug it in and turn it on, and it just works. This might seem like a pipedream, but it’s not.

A fine example of the evolution here is cleaning robots.  Roombas, at least for me, damaged my view of what cleaning robots could be. Maybe newer models are better, but I found myself spending a lot of time recovering it from being stuck, rearranging my furniture to suit the random algorithms they use to explore their surroundings, and even more time cleaning the robot than it spent cleaning the floor. It was just too stupid to be useful. Last year I finally cracked and got a Neato (who are in no way paying for this plug) and seeing that thing, the first time it’s been run in my apartment, crawling back around the inevitable mess to the charging station with no problems was one of the most impressive things I’ve ever seen. Watching a Neato plan how it’s going to attack a given area is almost creepy, as it uses SLAM to work out where it is.  This makes all the difference, and the result is something that requires almost no monitoring except to ensure that the dust tray gets emptied.

Extending this idea to the home at large I have come to the conclusion that rather than attempting any centralised control mechanism a better approach is to make more of the home responsive to the stimuli that is under the existing control of the inhabitants. For example, most lights could be smart enough to maintain an appropriate equilibrium for the time of day and get overall target intensity and colour temperature cues from whatever lights are directly under your control and the natural light for that time of day. Combine that with automated blinds or curtains that operate based on relative indoor/outdoor light intensity and you would get very complex behaviours which can follow what you want to happen without explicit programming.

In that case every light becomes a relatively dumb but self contained entity. The art is going to be creating entities of the right highly simplistic behaviours that in combination lead to the desired sort of apparent intelligence emerging. Then instead of programming a system you just put the right “personalities” of lights, blinds and so on together and they will just work. This does mean the home will do things you didn’t ask for, but this way allows you to be open for serendipity, without which your home just isn’t smart at all.

If you absolutely must control everything then make sure you have a FriendlyHome.

Nigel Birkenshaw runs Atomirex.

also of interest