Slashdot Mirror
The Sentient Office Is Coming
Roland Piquepaille writes "In this article, the Economist explains that "sentient computing systems are likely to be everywhere within five years -- listening and watching, and ready to anticipate their users' every need." "By adding sensors to today's computing and communications technology, sentient computing seeks to take account of a machine's environment in order to make it more responsive and useful. Sentient computing systems are always on, ubiquitously available, and can adapt to their users. In short, they seek to become real help-mates." ACM TechNews also wrote an analysis of sentient computing: "Challenges to sentient computing include the seamless integration of wireless networks, the spread of sensors throughout products and the environment, the accurate provision of location data, and the ability of sentient systems to merge vast volumes of widely disseminated data and customize its delivery for users. Other problems researchers will have to tackle include scalability, the development of cooperative file systems, and sentient applications' ability to find screens and network devices in close proximity to users." And of course, there are privacy concerns... Check this summary for additional details."
I am not sure I want a bunch of Clippy's running around in my everyday world second guessing what I want to do. Rather, I would prefer technologies that can do what I want when I invoke them. For instance, standing in your kitchen talking to your SO about a vacation, you say, "computer, find me the best airfare and hotel rates in say, San Francisco for the weekend of the...." The computer would then list those for you. I don't want my fridge saying "it looks like you are out of milk, I therefore ordered 1gal of milk for you" when I am going out of town for a week.
Visit Jonesblog and say hello.
Or something like it? The last Great Computing Hope was truly photorealistic graphics in games. I remember reading this in 1990. Well, 13 years have gone, and in-game images still don't fool me one bit.
Or will we be getting to this sentient office in our flying cars, eating a full course lunch that comes in a pill, and going back to our houses under the sea? Maybe I'll get my robot maid to make me a martini, for retro's sake.
...is that the concept assumes that office dweller knows what she wants.
I'm not being facetious. I work in an office, and I can't think of anyone who's particularly well-organized. The people who do their jobs well have good job habits, not a rigid system that an adaptable computer system can learn to predict.
The basic problem with anticipatory decisions by computers is that, if it offers something the user wants, it's accepted; if it offers something the user doesn't want, it's not just rejected, it's an irritant, an interruption. The cost of being wrong far outweighs the benefit of being right. Like branch prediction in the CPU, it has to be right far more often than it's wrong to be of practical value, and human behaviour is far harder to analyse and predict than computations.
Anyone who loves or hates any language, platform, or manufacturer, doesn't know what they're talking about.
But, wait, didn't they make this prediction back around, oh, 1966 or so? Nifty theorem provers would unlock the power of the computer for real Artifical Intelligence? No, actually, it was predicted even earlier than that, by no less than Turing. He figured we'd have machines capable of passing his "imitation game" test by the end of the 20th century.
Wrong on all counts. Speech recognition software still requires training and it's clumsy to use. Contents filters (as now mandated for libraries receiving federal funding, thanks to the oh-so-technically-savvy U.S. Supreme Court) still can't reliably tell the difference between breasts as in breast cancer and breasts as in porno. And the AI crowd is still grappling with things like knowledge representation schemas and semantic networks.
IMHO what we will most likely see are systems with huge lookup tables and canned procedural responses driving complex state machines, not flexible systems capable of introspection or foresight. It might even begin to exhibit what the philosophy/cognitive science crowd likes to call "emergent properties". It may even begin to become useful, but it most definately won't be sentient.
I have to admit, though, it would be nice to able to ask my house AI to list my appointments for the day and assemble a personalized news report from the wires while I brush my teeth and get dressed. But I trully don't think that'll be a reality until about the time I decide to pack it in and retire, if then. And then I won't really need it, or even care.
Pfft! They promised us flying cars and video phones, too, and I haven't seen any of those running lately, either.
Predictions like this are all fine and good, but we have a huge backlog of technology that needs to be developed. For instance, flying cars. We've been promised goddamn flying cars for years now. I want my flying car!
So before working on sentient offices, I suggest some extra time is devoted to such projects as flying cars, the paperless office, the helpful computer, and Duke Nukem Forever.
I hold out hope for all but the last.