Lately, there have been a lot of folks getting into the “AR excitement”, especially with the iPhone SDK becoming available (i.e., all kinds of folks _have_ to have an AR demo on the iPhone, even though the camera sucks and you can’t even distribute an app that uses video legally because the SDK doesn’t support it and it’s not “legal” to reverse engineer unsupported APIs). ”Sekai Camera” has gotten a ton of press, for example. As did “Enkin” before that (a mockup of an Android app on a mac, pre-Android phone release). Various companies have “point and know” kinds of technology, where the pitch is “using GPS and orientation information, combined with our vast wonderful backend database, you can point your phone at things and learn what they are.”
The problem, of course, is that these are really hard problems, and all of these systems only kinda-sorta work, even in their restricted demo modes. Can I really point at that doggy in the window (as the google folks suggest you’ll be able to some day)? Certainly not now. And, most likely, not any time soon! Could I point at the shop? Perhaps. At the items in the display case? Not likely.
The issue, of course, is that most of these so-called AR applications are more alluring than real. One huge problem is that the amount of information needed to deliver on the hype is mind-boggling; it’s the scale of information that will never be available in a closed system, in just the sort of system most of these demos are pushing.
And, like the VR hype before this, and the AI hype before that, the worry is that (since none of these systems will do what they purport to do) the overhype will kill the potential industry and possible market. There are companies who are tackling more modest problems, but they don’t get the PR and can’t create web memes because they aren’t as flashy. That’s shame.
Because I’d hate to see AR creep back into the lab with it’s tail between it’s legs. None of us who’ve been working on AR for decades want it to be the next “Big AI”.
2 comments
September 26, 2008 at 3:03 pm
The “mind-boggling” amount of data you want is already there.
Almost every product sold in the western world has a barcode.
Cross-referanced with the stores database, and possibly something like google product search, and you have everything you need.
All we need from a technical standpoint is to have barcodes visible for AR devices to see.
The rest is “just” co-operation and establishing of standards.
You are right that a closed system stands little chance of success, but from a technical standpoint of a lot of the “AR Dream” is quite possible.
Especialy as the “point to look up a product” is just a tiny aspex of the overall usefullness of a decent AR device.
September 26, 2008 at 6:05 pm
I totally agree that some of the data we need is there; yes, we can identify products, IF a device can read the barcode. And, sure, it’s possible to know what’s in a store IF you can get access to the store’s databases.
However, those are some huge IF’s. And some huge leaps of faith. For example, name one store that keeps a SPATIAL record (computerized, not just implied or ad hoc) record of EXACTLY (3D coordinates, please) all those items are. You might be able to find out what is in the store, but not where it is. How do you know what is in the window?
Some folks think RFID will be the answer, but it’s been field tested and largely has failed. Big companies have eschewed putting RFID tags on products because of the cost and lack of benefit to them (as opposed to benefit to the stores).
Do some exercises for yourself to try out how big a leap you are making:
- go to a store and look in the window, and see how easy it is to see the barcodes on the display items. Bring the best DSLR camera you can get (e.g., I have looked at pictures taken with my Nikon D300). Zoom in; sharpen up those images with Photoshop. Even IF you can see the barcodes, can you read them? Could a computer? Be honest.
- Now, consider some of the “hyped” scenarios being put forth by folks at companies (like the guys on the google blogs). Stand back from that window a moderate distance. Look at that stuffed dog toy in the window. Connect all the dots in the system you image.
Sure, “everything” is possible with enough assumptions and “a small matter of programming.” Hypothetically, standards and co-operation can solve lots of things. But, when will these standards be here? Why would they be created? Who is paying, and why?
I do completely agree that this will happen. The point of my post was that it’s not as close as the demos and emerging hype would have “the common folk” think, and it is just this kind of “too soon hype” that the VR and AI communities engaged in.