Reflections from the land of defragging

Today’s New York Times carries a story called “Seductive Electronic Gadgets Are Soon Forgotten” about how people generally are defeated by the complex user interface of the widgets that they purchase.  I spent about 15 hours over the weekend checking, repairing, and defragmenting the hard drives on desktop and laptop PCs (both about 1.5 years old and apparently overdue for some treatment).  What if there is a fixed percentage of human life that people are willing to devote to reading owner’s manuals, learning new interfaces, and remembering which button does what?  If true, that places a limit on the growth of the electronics industry.  Until there are natural speech interfaces to gadgets (“hey, this DVD is in 16:9 mode; compress the image please” to the TV; “turn on all the lights” to the house), how can people buy more?


A good example of this is navigation systems for cars.  You’d think that everyone would want a moving-map GPS in his or her car.  Certainly it would be nice to say “I’m going to Joel’s house” and have the navigation system pull his address from a cell phone or PDA and then provide voice and map guidance.  What if, however, it is a 10-minute trip and you’re pretty sure you know how to get there and you don’t have that natural speech interface?  Are you willing to spend 2 minutes programming today’s clumsy GPS units with Joel’s address?  Will you accept a 20% increase in your travel time to avoid the risk of getting lost?


Are we ready for another boom in Artificial Intelligence research, this time funded by car and gizmo companies?

24 thoughts on “Reflections from the land of defragging

  1. “Will you accept a 20% increase in your travel time to avoid the risk of getting lost?”

    No. But then, I already admitted not owning a car and fitting these gizmos to my bike would be tricky at best.

    I am sometimes amazed at the amount of gadgets people seem to need nowadays. How on earth did man ever find his way, or keep track of appointments, or make a photograph, or even switch on the lights.

  2. One of the cool features of the Palm-based Garmin iQue 3600 GPS is it’s integration with other Palm apps.

    For example, if you have Joel’s address in the Palm address book, you can just select the “Route To” command and it’ll tell you how to get from your current location to Joel’s house with visual and voice commands. If you make a wrong turn along the way it’ll recalculate the route and tell you how to get back on route.

    It’s a slick system and works well — as long as the basemap and routing information is accurate.

  3. I think we need a universal wireless power source to really take advantage of all the gadgets. I’m tired of charging and charging and putting batteries in everything… I’ve heard NASA were able to keep a model plane airborne indefinitely by shining a laser beam from the ground on its photo cells. We need something like this for everyday toys and it has to be pretty safe…

  4. For a wonderful article about the frustrating complexities of digital devices, see this New York Times piece from April 28, 2002:

    ===========================

    Comforts of Home Yield to Tyranny of Digital Gizmos

    By KATIE HAFNER

    Last Christmas Eve, just as Lynne Bowman was preheating her oven to roast a turkey for 15 guests, her daughter accidentally brushed against one of the new oven’s many digital controls.

    “We heard this `beep beep beep,’ ” recalled Ms. Bowman, a 56-year-old freelance creative director who lives in Pescadero, Calif., “and no more oven. After that, we couldn’t get it to work.”

    Ms. Bowman’s husband, an engineer, was unable to fix the problem. Nor were any of the assembled guests, half of whom were also engineers.

    Desperate, Ms. Bowman resorted to the small, simple 1970’s-vintage Tappan electric oven in the guest house, which worked like a charm.

    Of all the forces that permeate daily life, perhaps nothing has become more of a tyranny than the bits and pieces of technology that are meant to help one get through the day more easily, but instead are a source of frustration.

    Relatively simple devices that were once controlled by twisting a knob or pushing a button are now endowed with digital commands that can take hours to master.

    Many televisions, inextricably joined with the VCR, DVD player and 500-channel receiver, are now impossible to turn on or off without first scanning a cryptic array of choices on any of several remote control devices.

    ===================================

    http://www.longdistanceconnection.com/artdigitalgizmos.html

  5. I think either the devices are difficult or some people are becoming increasingly dumb so they can’t follow simple instructions.

  6. 1. Gadgets are hard to use. It’s hard to type names into my cell phone. And it’s hard to write graffiti with the Palm Pilot. UI and I/O are huge problems. Voice is nice. But it’s too much of a broadcast. Computer lip reading doesn’t work, yet. Brain implants?

    2. Why not go further and build a car that drives itself? Why not go even further and just say “beam me over, scottie!”? Hell, why not go off the deep end by building brain implants that allow you to communicate and interact virtually with your friends?

  7. “…Why not go further and build a car that drives itself?”

    Good idea.

    Do you know that Congress has mandated that in the near future 30% of the US Army’s ground battle systems must be unmanned autonomous vehicles?

    And there’s something called “The DARPA Grand Challenge” (http://www.darpa.mil/grandchallenge/index.htm) with a million-dollar award for the team that fields the first vehicle to complete a designated route within a specified time limit.

    And didn’t I read something on CNN the other day about a researcher that hooked a monkey up to a robotic arm that could be controlled by the monkey’s brain-wave patterns?

  8. As the proud owner of a car with a GPS navigation system (a Honda Accord 2003), I should first point out that simple voice interfaces are here. For instance, I can tell my car “Go Home” and it will calculate a route home. That’s the only case that is that simple, but it’s, by far, the most important one. “Find the nearest gas station”, “Find the neareast ATM” and “Find the nearest Mexican restaurant” work, too.

    Another interesting optimization is the ability to go to a place by its phone number. Specifically the commands: “Find place.” “Phone number.” “617 492 7777” “Search.” will convince my car to calculate a route to the Bisuteki Japanese Steak House in Cambridge, MA. This would be more useful if the point-of-interest database were more complete and up-to-date, but it can still be helpful.

    I’ll admit that the voice interface is not close to natural speech (and voice commands I use most frequently are for changing CDs), but I think it is impressive and it demonstrates the interesting short-run possiblities of menu-based voice interfaces.

  9. Had some interesting experience with Hertz “Never Lost” GPS system. In Orlando’s Hertz car lot, to my great surprise, the system could not find where it was. It did find it way after I drove the car out on the streets. In Detroit, on many roads, one has to make past the crossroad and make a U turn + a right turn just to turn left. This defeated the system again. It asked you to turn left right away, which is against the traffic. Bottomline is you can only count on it for a general direction, up close to the location, your eyes are the most useful equipment. This goes with other gizmos. Marketing people never want you to wake up to that – they are selling not a product but a dream …

  10. I may be weird. Why? Because becoming lost on any trip is the most fun part of the journey. You see more, learn more and experience more than expected. Plus it gives one an opportunity to exercise those brain cells. It is amazing how many people are unable to use a map, converse with people to ascertain directions, pay attention to surroundings to keep from traveling in circles.
    As you can tell, I am not part of the market wanting any type of GPS directed travel. Of course I may eventually because a favored pastime for me is figuring all the ways to use a device in unintended ways and its limitations.

  11. Alex, that story you posted about “even engineers couldn’t figure it out” has it all wrong.

    The reason engineers can’t figure it out is because it isn’t designed to be used by engineers! Had the oven had a TCP interface with telnet server you could log in to and use commands like: “oven clear”, “oven cook -mode=fan -temp=190 -time=25m” (and ofcourse an “oven –help”!) nobody in the room would have had a problem with it.

    Same for the Creative Nomad Jukebox Zen NX that just arrived on my desk. On my PC, I have all my files in either “music/albums/artist/album/” or in “music/singles”, with playlists for the latter. This is inmensely easily to navigate and find what you are looking for. The Zen insists on sorting by either artist/album|song[/song] or album[/song] based on the ID3 tags. This means that if I want to get to an album, I have to either plow through scores of one-hit-wonder artists from my singles collection, or go straight to albums, where I am presented with all the “albums” in the ID3 tag from my one-hit-wonders (“Now! This is music 5346”, etc).

    This means the only option I have is to use a sledge hamer script that blasts the album name to “Singles” for these. But then I still have no way of going from a small list of artists to their respective albums. I will still end up with one long list of all albums.

  12. I’m still waiting for the returns on the first boom in AI research. So far, all we’ve gotten is a computer that’s really good at chess (whose programmers may have cheated to make it that good, btw). And you could argue that chess problems are not really AI problems anyway, but that’s another post for another time.

  13. Some friends of mine half-jokingly define AI as “the stuff we don’t know how to do yet”.

    Depending on whose definition you use, AI is doing dozens of things people use every day. Examples I can think of offhand are running voice recognition systems for phone-menu systems, finding the fastest route to drive to the store, finding the cheapest way to fly to Miami, translating a French web page to English so you can read it, and watching security cameras to report traffic accidents.

    Of course, people have trouble calling these “AI” because in the movies “AI” means “human brain in a computer”. The fact that they’re working systems today somehow makes them boring.

    Yeah, we’ve got really good chess computers. But don’t let that distract you from the actual accomplishments of the AI community.

  14. I think the key is invisible interfaces. How about a mapping system which grabbed addresses from my email and put stars on the map as I drove. Scenario that happens a lot for me is that I’ll get email, either as an event announcement or from a friend, and drive to that address. Addresses are fairly easy to parse from email, we can already show the map centered on the car, usually I can find a neighborhood without turn by turn instructions, and the data is sparse enough that you could probably just put a “*” on the map for every address I’ve ever gotten, maybe color coded for how recently I’ve gotten it.

    Now if I can just figure out an easy way to parse TIGER data, I can build this puppy for my laptop myself.

    The major problem I see in the computing industry right now is that everyone’s talking about making new interfaces. We already have tons of good interfaces that we’re quite comfortable with, from doorknobs to snooze buttons to email. Start using the useful information from those first, we’ll worry about adding more complexity later.

  15. Chess computers completely lack opening strategy and are even more helpless in the endgame. A chess computer makes first 10-20 moves from an openings database that humans put together. This is done in order to reach a position that is comfortable for evaluating (e.g. with lots of tactical possibilies). If you take the opening database out computers become really stupid and raw calculating power doesn’t help. The same happens during the endgame when there are just few pieces left on the board: machine looks at a precomputed database (which is btw hundreds of Gigabytes) of forced winning endings and tries to reach that position. This is definately not the way humans play this game.

  16. “This is definately not the way humans play this game.”

    I think you’re still stuck in the “AI = human brain in a computer” rut. Emulating the human brain is one part of AI, but certainly not all of it. You can also make a system which emulates human responses (with a different mechanism), or make a system which behaves logically.

    Most chess computers seem to be designed foremost to play winning chess. Does it really matter that they don’t play the same way humans do, or that humans gave it certain precomputed data to work from? My AI textbook says this is a perfectly valid form of AI. I don’t see why any of this matters, either: the problem space is well-defined (win at chess); “emulate the human brain” is quite intentionally left out of the problem.

  17. the problem space is well-defined (win at chess); “emulate the human brain” is quite intentionally left out of the problem.

    I think this approach worked only because chess is a relatively easy problem to solve by using brute force computation. Another game “Go” has even simpler rules than chess but computers failed miserably so far. Humans use some kind of universal mechanism that allows them to make decisions based on intuition. Someone has to figure out that mechanism and simulate it on a computer that’s when we’ll have “true” AI.

  18. Dan, I didn’t even think of it that way, but it is true!

    Especialy now that more and more devices have menu systems, as opposed to many fixed-function buttons, to operate them, why not simply support both ways on my MP3 player?

  19. I am trying a few things with natural language interfaces. Please check out http://www.EllaZ.com One of the things we do is display images in response to conversation, including a number of Phillip’s photos.

    “Ella” can deal Blackjack, retrieve WordNet lexical data, tell jokes, display books, do word math, and so on.

    The real purpose of this post is to track down Phillip and see if he will permit us to use some of his photo’s on a CD-ROM version of a new “AIBush” game we plan to distribute.

    The AIBush game features natural language access to more functions via web services, plus chess and a “Reelect Bush?” game where the player acts as an advisor. The http://www.AIBush.com site should be complete in another day or two.

  20. I have a friend who worked for a company that had a NASA contract for building telematics interface software, with the eventual goal of commercializing/co-developing that technology to work in cars. He left that company to go work for a company which builds hobbyist robots and is now getting more lucrative contracts by selling their technology as an OEM vendor.

    If car companies are working on telematics, they’re doing it in-house, because they don’t seem to be buying the stuff NASA contractors put out. But then, I’m not in the industry, and so I’m likely horribly wrong about the actual economics and business side of things there.

    OnStar seems to be a mediocre achievement in terms of rollout (is anyone other than a few big American car companies including OnStar systems in their cars?), and a gigantic failure in terms of a business, but again, this just my impression of what I’ve seen.

Comments are closed.