Do you revile the concept of "intelligent homes"? I do. And I'm not some "luser" geek-snubbing Poly Sci major who's never burned his finger with a soldering iron. No: When I was 15 and the other guys were out getting dates, I wired up my room with momentary-contact switches and latching relays (we didn't have transistors in my day) [that was a joke] that made it possible to control the various lights from the easy chair, the bed and the door. Many intelligent home application proposals are loopy and even dangerous. (For reasons that escape me today, I regarded this as a crucial architectural enhancement to the space.) So I understand some small part of the home automation urge, okay? In case it's not obvious, I intend to go off on a rant here. Then I'm going to explain why I'm wrong. I tell you this now so you don't suffer intellectual whiplash. With that level-set, let's look at some proposed applications that pop up year after year in the intelligent home community.
I can call my oven from the office and turn it on so it bakes the casserole I prepared before leaving for work. That way I have a nice warm dinner waiting for me when I get home. Do I really want my oven on when I'm not in the house? Particularly when any smoke it emits will cause my alarm system to call the fire department, which will break down my door and be unamused to learn that the smoke's existence is owing to the fact that I haven't cleaned my oven in over a year? Worse, it could start a real fire.
My light switch "learns." Some researchers have tried to make a light switch that "learns" based on the circumstances under which you turn it on or off. For example, if your tendency is to turn lights on when you enter a room and turn them off when you leave, then the switch might notice this and begin to act on it. Unless it's daytime, in which case it might notice that you tend to leave the lights off. Unless there's a daytime thunderstorm, in which case it might notice that you turn them on. Unless someone's sleeping in the room, in which case you leave them off. Unless, except, on the other hand, etc. The ironic part is that I'm perfectly willing to make the decision myself (it takes about a half-second) and communicate it directly to the system using a simple, intuitive "wall switch" interface (a metaphor borrowed from the GUI radio button). Meanwhile, as my colleague Andy Fano likes to point out, the system is thinking something like, "Wait! Wait! Don't tell me! I can figure out what you want!" In short, this is a very tricky AI problem--almost as tricky as it is irrelevant.
My front door recognizes me. The idea here is that a camera in the peephole uses face recognition software to decide whether you're the owner and, if so, unlocks the door. One problem is that face recognition is still a fairly chancy business (I've done a bit of it)--false negatives and -positives are still common, even with small databases. Another problem is that it'd be easy to print your picture (perhaps from your Web site) and hold it up to the peephole--instant burglary. And if the door wouldn't open for you for some reason (maybe the lens got dirty or the sun is in an awkward place), you're back to a key--which means you'd better not forget your key, which means there's not much point in the face recognition system. (Note that you could do something like this much more effectively using an RFID chip mounted inside your hand and a reader built into the door knob. But that probably wouldn't be "intelligent.")
I can talk to my room. You are lonelier than we thought. That aside, the vision here is that you can speak commands in order to control lights, television, drapes, etc. The problem is that even with a small vocabulary, individual training, and a carefully placed microphone, speech recognition accuracy is still annoyingly less than 100%. If the microphone is built into the ceiling (say) rather than close to your mouth, or you have a cold, the accuracy will drop significantly. And what if the television is on while you're trying to speak a command (in order to, for example, turn it off)? Or the TV accidentally speaks a command ("Call the police!")? What if you have chattering guests? Despite their well-known deficiencies, there's a reason we use conventional remote controls. (Now, if someone wanted to tackle the problem of remote control proliferation, that might actually be useful.) The oddest example of intelligent rooms came from the MIT Media Lab, some of whose researchers once suggested that, in the house of the future, you could tell a joke in an empty room and the walls would laugh. (No kidding.)
The seeds of their own destruction. I seem to remember an old Perry Mason episode in which the murderer used a radio transmitter to turn off a gas fire (he'd secretly attached a small receiver and a solenoid) then turn it on again. (Think of him as the original remote hacker.) The unburned gas went on to kill whoever was in the room. This raises an inescapable problem with smart houses: the smarter they are, the more vulnerable they are. Example: I was recently in a spectacularly luxurious intelligent house. In addition to a central panel from which all 100+ lights could be controlled, it had a similar panel from which the water to three dozen toilets, sinks, and fountains could be individually turned on and off. Imagine the sheer hacker joy of breaking into this house's systems and shutting down its toilets during a New Year's Eve party. Of pumping unstoppable, high-volume 2Pac out of the house's four surround sound systems. Of reprogramming its wall switches so they simultaneously summon the fire department, the police department and (what the heck) the IRS. The house itself was wonderful and its systems were a lot of fun, but it was very, very vulnerable.
I am, however, obstructively cynical. It turns out that there are intelligent home applications that make a lot of sense. They differ in kind from the examples above in that they focus primarily on sensing and reporting rather than control. Most of them revolve around helping the elderly (which is to say most of us, eventually) "age in place." Here's a suggestive but by no means exhaustive list.
- SensorMat (I've heard of this one over the years, but I can't find a link to any research so I made up the name. Sorry.) This is a piece of carpeting placed in a high-traffic area. Built-in sensors track your progress as you walk across it. A watching system reports to your doctor or a relative if your gait changes significantly--which might indicate an alteration in your "pain profile" or increasing problems with balance.
- CollapseCam (Developed by Accenture Technology Labs. Again, I made up the name, this time because I don't like the official name ("Activity Monitor" ).) This is a camera-based system that senses whether you're lying on the floor. Since that's not a common position for most adults, it's likely to indicate that you're in some sort of difficulty (like maybe you can't quite reach the Yahtzee die that's rolled under the couch). CollapseCam could be set up to alert a relative or other source of help.
- Inertia sensor This is simply an electronic scale that tracks how long a person sits on a chair or couch. If the duration gets out of hand, it signals for help.
- Motion sensors We're used to thinking of motion sensors in terms of burglar alarms, but there's no reason they can't be used to monitor a resident's movements as well. They'd report to a system that would maintain minimum activity thresholds and signal for help when they were violated.
So, in the end, I'm wrong. True, many of the intelligent home application proposals are loopy and even (in some circumstances) dangerous, but the problem of an aging population has stimulated a whole new research area that looks sensible and even exciting...so much so that it's tempting to get out the old soldering iron and maybe do a little damage to my parents' Florida dream home.