I'm writing this from ISIT (Int'l Symposium on Inf. Theory) in Austin. In fact, I'm sitting in on the session on for "Coding for Memories", a topic I did some work on last year and could see returning to (if I could figure out a problem I thought was sufficiently fun theoretically and sufficiently close to practice). I'm at the conference to give the talk on "Tight Asymptotic Bounds for the Deletion Channel with Small Deletion Probabilities" (work with Adam Kalai and Madhu Sudan; here's my web page version which might not be entirely up to date) tomorrow. This is one of those massive 8-parallel session conferences. For the record, I count about 50 people in the room, and my guess is this is one of the less popular sessions. I'm finding the talks interesting, though I'd admit that perhaps they're for a specialized audience. The pros-and-cons of the multi-session large conference.
I just got in a few hours ago, and I'm afraid this conference is a "fly-by" for me -- in and out as quickly as possible. (Sadly, from Boston to Austin, there doesn't seem to be a convenient direct flight these days, so even a fly-by is a 2-day affair.) Too much recent conferencing-- Bertinoro/STOC/EC -- meant I was eager to minimize this trip. I'm the first talk in my Friday early afternoon session, and in order to get a flight out to Boston Friday, I have to leave before my session finishes!* I feel a little bad about that, but I'd feel worse about not being home until late afternoon Saturday.
I'll enjoy my time here -- plenty of talks look interesting**, plenty of people to say hi to, and plenty of other work I can get done while at a hotel or in an airplane. But it does make me wish perhaps there was another way. Giving a talk by video? Conferences in an electronic avatar space? I know classes that are recorded and put online generally get smaller attendance, as students don't come to class as often knowing they can catch a stream of the lecture later -- would our way of doing conferences change significantly if people could "go" to conferences electronically without actually traveling to them? (The experience of STOC/EC, where I went home at the end of the day, was vastly different for me than other conferences where I'm there all day. There are pros and cons, but in general, I'm at a stage where I like going home. I'm sure that will change as my kids grow up and tell me they'd rather I stay away.)
Until such conference changes occur, though, I'll continue to be timing my flights carefully.
* Thanks to ITA software, I also found I could leave later, fly west to LA, and then take the red-eye to get to Boston Saturday morning in plenty of time for breakfast. I almost did this -- more frequent flier miles! -- but the ticket was a couple hundred dollars more. And my wife reminded me I'm not exactly at my best after a red-eye.
** For the CS crowd, Zhou/Bruck had a paper on extracting independent uniform random bits from a Markov chain efficiently -- a TR version seems to be here. This is an old problem -- Manuel Blum, for instance, had a paper on it. In fact, my first lecture in my algorithms/data structure class is on the simpler, earlier-studied problem on how to get unbiased bits from biased bits, as I described in this post a while back.