Monday, May 23, 2011

You want to attend this conference? Become an optimization expert!

Imagine a conference so big, so sprawling, that navigating it requires you to draw multiple decision trees and redraw them on the fly as needed. Why? The conference is spread out across a convention center the size of four city blocks, several neighborhood hotels, as well as some hotels a few miles away. Just walking from one room in the conference center to another takes 5 minutes; a walk to a neighborhood hotel takes 10 minutes, and to a faraway one, as much as half an hour.

And when you get to your destination, you discover that the organizers picked too small a room for the panel, and there's a line to get in the door*. You can only get in if someone leaves, and no one is going to leave because they are as interested in the panel as you. The speakers don't use a microphone, so you can't hear a single word from outside the door.

Imagine all this, and -- that's right! -- you'll get SXSW.

People wait to take an escalator down from the 4th floor of Austin Convention Center

To get an idea of just how many people were at SXSWi, consider that this amorphous line is just the people waiting to take an escalator down from the 4th floor of Austin Convention Center (and some getting off the escalator in the opposite direction). More pictures from SXSW 2011 are in my photo gallery.

Congratulations, you have walked 10 minutes for nothing. It's time to pull out your handy decision tree. Say node A represents your preferred panel. It is linked to nodes B, C, etc., that designate the next most interesting panels to go to in case panel A is full. Each node is assigned a weight according to its desirability. The edges that connect the panel-nodes in turn should be weighted according to walking times required to get from building X to building Y. And if you want to make the problem more realistic, you could make those weights depend on pedicab availability, or your willingness to pay for a pedicab, as well as a probability that the free sponsored shuttle will be waiting outside the hotel at that moment. Then voila! Solve the optimization problem, and you know where to go next!

Is there an app for that? I don't know, I still don't have a smartphone. (That alone should disqualify me from SXSW attendance, I suppose.) I know SXSW has an official app, but I doubt it solves optimization problems. If it doesn't, should I spawn yet another clone to write such an app (joining the league of all my clones in parallel universes that are writing other applications I've dreamed up, while I'm toiling at my job)? Then, and perhaps only then, I would have a real reason to come here again. After all, as everybody says, the real purpose of going to SXSW Interactive is to schmooze with other entrepreneurs and wannabes. Panels be damned.

Cloaking geeky topics in cutesy terms

Maybe panels should be damned. Most panels were for internet marketers, since that's what most SXSWi attendees are, and marketing is not that interesting to me. Software developer panels were dominated by mobile development, a perennial hot topic; unfortunately, I haven't done any mobile development and I don't know if I'll ever get a chance to do it where I could get paid for it. There was a good presentation on the basics of semantic web -- a loose set of methodologies that let us mark up the meaning of the text, instead of just its structure, in an HTML-like way. This allows computers extract meaning from texts they "read". The panel would have been even better if it hadn't tried to hard to achieve mainstream appeal by drawing parallels between semantic web and dating. But even renaming Linked Data Principles "The Bro Code", and urging people to "get your data a date" (i.e. link it to other data on the web) did not attract more than 10 people into the room. What the presenters did right, though, was to illustrate the notion of a semantic triple as something we routinely create in our daily lives when we fill out forms. This, and not sketchy dating analogies, would make the concept of semantic web more interesting to people.

A patient information form illustrates a semantic triple A patient information form illustrates a semantic triple. More pictures from SXSW 2011 are in my photo gallery.
A triple: the person filling out the form is the subject, field labels are predicates, what you put in the blanks are objects An subject-predicate-object triple: the person filling out the form is the subject, field labels are predicates, what you put in the blanks are objects. Thus by filling out a patient information form you are creating triples such as {Joe Schmoe, Home Address, 1000 Main Street} or {Joe Schmoe, Date of Birth, 1/1/1900}.

General interest panels on emerging technology, such as gesture interfaces, or "internet of things" weren't all that worthwhile to me because I read a lot on those topics, and the presentations didn't add anything to my knowledge. The curse of being ed-yoo-ma-cated.

So after a while I lost motivation to pick the best panels and brave the crowds to get to them. Instead I spent big chunks of time just milling around the convention center, parties, and satellite events, not going anywhere, and cursing myself for wasting the time I'm paying for with my own precious future time, that is to say, those volunteer hours I'm obligated to put in. Speaking of which...

Volunteering shows its dark side

After I had an easy time volunteering at SXSW in 2011, I expected something similar this year. Yet volunteering showed its sharp claws this time. As before, I spent most of my time at Film Venue. Most Film Venue volunteers perform a vague function of "line management", which is to say, they stand around and see that the audience lines up in an orderly fashion: one line for SXSW film badge holders, another for film pass holders, and yet another for those with individual tickets. Most movies don't get a lot of audience, so there isn't much standing around to be done. However, on my first day the theater manager told us to take our positions an hour and a half before the movie started. We were supposed to stand at attention the whole time. Needless to say, standing on your feet while 6 months pregnant isn't a picnic. It's also excruciatingly boring. I would rather have spent that time moving furniture (well, lightweight furniture) than standing still. So I went to the theater manager and told her I couldn't do this while pregnant. She said she didn't realize my condition (what, did she think I carried a basketball under my shirt?), and told me to go home for the day. The next day the management agreed to transfer me to another theater where I got a desk job. For the rest of my volunteering week, I sat at a desk and answered customers' questions. That was much better.

All in all, I should try coming here one day as a wannabe turn-your-side-project-into-a-startup entrepreneur. I heard those people get the most out of SXSWi. Until then... I don't know.


* That was the case with "Agile Self-Development" panel. Having been introduced to agile software development concepts at my recent job, I was curious how this would apply to self-development. Should I hold daily scrum meetings among multiple facets of my personality?

Wednesday, May 04, 2011

SXSW 2011: People as Peripherals: The Future of Gesture Interfaces

"People as Peripherals". The title of this speech conveys unease about a future where humans are little more than input devices for our computer overlords. Not surprisingly, presenter Lee Shupp segued from gesture interfaces to brain implants, and from there to technological Singularity.

At the first glance, there is neither much to fear, nor great promise to expect from such gesture interfaces as Kinect, a Microsoft game console. Current interfaces suffer from the case of "gorilla arms": you have to wave your arms vigorously in big, sweeping gestures to make yourself understood to the machine. You are also limited by a small square of space where you need to stand so the computer would capture your gestures correctly. Even so, it's all too often inaccurate, if Kinect is any indication. It's a long way from here to detecting micro-gestures, such as subtle finger movements.

Lee Shupp

Lee Shupp speaks about gestural interfaces, brain interfaces and Singularity. More pictures from SXSW 2011 are in my photo gallery.

Not unlike at a science fiction convention, the audience pointed out plenty of other problems gesture interfaces will have to solve before they are seamlessly integrated into our lives. How would a gesture-driven plane cockpit respond if a pilot sneezes? How would such interfaces adjust for body language differences between cultures? For example, in many oriental cultures it's considered extremely rude to point your foot to anyone. Never mind the bugs -- the potential of well-implemented gesture interfaces can be equally disturbing. A guy in the audience expressed a wish for an interface that would understand sign language. He can sign much faster than type, and he'd like to "text" while driving without raising his hands from the wheel. (I sure hope for the sake of the humanity that his wish won't come true.)

But before we can even make sensors that understand sign language, there are more basic problems to be solved. As a person in the audience pointed out, current interfaces require that you come to them. You are supposed to stand in front of the machine and wave your arms at it. That doesn't integrate well with our daily lives. However, I saw this Technology Review article, Talking to the Wall, about an experimental technology that lets you turn any wall in a building into a touch-sensitive surface. Now that surely has a few killer apps in it.

Brain interfaces are still in a rudimentary stage too, says Lee Shupp. So far brain implants haven't done much more than allowed people control a cursor on the computer. There are serious obstacles to their adoption. To connect a brain to a machine you have to drill holes in the skull, and sending thought commands requires concentration, which is hard for humans to achieve in the multitasking world. Finally, Shupp says, if people can't read people, how can computers? For that matter, if computers can read our brain signals, does that mean we can't lie anymore? To the guy who asked that last question, Shupp recommended "The Truth Machine" by James Halperin, a SF novel where this is addressed.

Lee Shupp's vision of transhumans

Lee Shupp's vision of transhumans. More pictures from SXSW 2011 are in my photo gallery.

Despite these nontrivial problems, he believes brain implants will take off. Already 80000 people worldwide have them. An informal survey of the room shows that most people here think we will be using brain interfaces in 50 years. At some point brain implants will likely augment our intelligence, and we'll on the road to Singularity. And then, if this slide correctly reflects Shupp's vision of transhumans, we will spend our time with our brains plugged in directly into simulated medieval worlds. Swords: the original gestural interfaces. ;-)

(Tangentially related, here is another take on Singularity, where the original popularizer of the concept, Vernor Vinge, discusses the concept with several science fiction writers.)