Archive for the ‘Research & Artificial Intelligence’ Category

in which cashiers are not as encouraging as they intended

Thursday, May 14th, 2009

Yesterday morning I defended my dissertation proposal. It went pretty well. I passed, which is really the main thing. Several members of my committee, and several other members of the audience, told me the presentation was good, which was a bit of a surprise–I thought it went fine, but didn’t feel like it was great.  I guess I’d lowered expectations by giving a series of really, really bad practice talks.  I didn’t finish revising the slides until 8:30pm the night before, and my first practice after that (to an empty room) was terrible, so I practiced a second time the night before, and it wasn’t until that last practice that I felt like it might go okay.

Anyway, my committee was all pretty positive.  Their primary feedback was that they wanted me to focus my scientific energy on one of the two aspects of my proposal (Motivation) and less so on the other (Maturation).  I fully expected and welcomed a conversation with my committee about how to pare down what I readily acknowledged was an ambitious proposal, and I’d felt as though I was stretching the Maturation material to give it equal weight with the Motivation material, so that feedback doesn’t really bother me at all.

Anyway, the two episodes I found amusing occurred as I was picking up refreshments for the presentation.  The Dunkin’ Donuts cashier who helped me carry the two boxes of coffee and more-bagels-than-god out to my car asked me if it was for an event, and I explain that I was giving my dissertation proposal defense.  She asked what in and I told her, and then she told me that she’d worked with another group in my department some six years ago.  I appreciated the friendliness and all, but…now you’re working at Dunkin’ Donuts.  Not exactly encouraging.  (I didn’t say that, of course, and it’s quite possible–probable, even–that she wasn’t a student, but rather a tutoring-systems test subject, but still…)

Then I went to the world’s most gimmicky grocery store (Big Y) for some juice.  I put my two bottles of fruit juice and four apples on the belt, and the cashier asked me how I was.  “A little nervous, I have a big presentation this morning”, I answered.  Then she asked me, “Would you like any grilling spices with that?”  WTF?  They had a display at the checkout, and I’m sure she had to ask everybody, but I had two bottles of juice and four apples. I answered, “No thanks, the only thing likely to be grilled this morning is me.”

So much for my career in the toy industry

Wednesday, April 22nd, 2009

First the Aibo, now the Pleo–Ugobe’s filed for bankruptcy.

(I’m not surprised that there’s no market for entertainment robotics, especially in this shitty economy, but I’m still disappointed, as I see that as one of the most plausible “practical” applications of my research.)

New Yorkers help a little lost robot

Sunday, April 12th, 2009

This is really cute.

Not stimulating me

Saturday, February 7th, 2009

Hey, Senators.  Let me give you one of those real person stories you love so much, about this economic stimulus bill you’re trying so hard to fuck up.  I’m a graduate student, researching the kind of technology that could drive the future’s economy.  Thanks in part to the economic crisis, I’m under-employed — I only have a half TA this semester, and no research funding.  This means I won’t have enough money to pay my bills, let alone drive the local economy by buying stuff.  But you just fully eliminated the spending most likely to support my research: NASA, National Science Foundation, and higher education funding.  So I’m not likely to be helping to turn this economy around in the near term.  And I won’t have as much time or support for my research, so the longer term doesn’t look so good either.  And when I do finish my degree, I’m likely to take my highly-educated, science-and-engineering, future-technology skills to Europe or Japan, where they’re investing in the kind of work I do.  Where I will help develop the technology to drive future economic strength, and spend some of what I earn doing so in their local economy.  And perhaps train the next generation of scientists and engineers, who will increasingly not be American, because we’re not investing in our primary and secondary education, either.  And it’s too bad because, the appeal of Europe notwithstanding, I’d mostly prefer to stick around here.  But you’re too damn shortsighted to steer this country out of the brain drain and general decline we’re hurtling toward.  In fifteen years, when the U.S. has lost stature while China, India, Europe and Japan have all moved ahead, don’t tell me you didn’t ask for it.

robots make stupid mistakes exciting!

Thursday, October 23rd, 2008

I don’t write much here about work/school, but I thought I’d share a little anecdote to give all of you a taste of how I spend my days.

Dividing by zero is a common mistake in programming.  You’ve got some variable, and you forget to take into account that it might sometimes be zero, and you try to divide something by it.  Often the result is “NaN” (not a number), and that result can get propagated through your program, causing something to break and not work like it’s supposed to.  In regular programming this is no big deal…if your program doesn’t work like you expect you start printing out the values of variables, you see “NaN” where you expected something like, say, 0.1, and then you track it back through the code to figure out where you’re dividing by zero and fix it.

But if the program is for controlling, say, the joints in a robot’s arm it gets a little more exciting.  Instead of your program just reporting a nonsensical result in a window on the screen, what happens is that instead of a slow, smooth movement like you expected, your $50,000 robot goes apeshit and flings its arm into some crazy pretzel shape as fast as the motors will go, and you have to leap up and hit the big red button that shuts everything down and hope that nothing was damaged.  That’s what happened to me yesterday.

28

Thursday, May 22nd, 2008

Breaking with tradition, I had a rather nice birthday.

I jokingly commented on Tuesday that the two things I wanted for my birthday were a Democratic nominee and summer research funding. Neither is in the bag, but Oregon gave Obama an absolute majority of pledged delegates, and this morning I got some unexpected promising news about summer funding. When my advisor asked me to come meet with him to talk about the summer, I figured he was more likely to read me the riot act than anything else, and I didn’t think it was likely that he’d pulled someone else’s funding to give it to me. Turns out a new grant has probably come through, and he will probably use it to fund me this summer — neither of these things is certain yet, but the both sounded fairly likely. The grant isn’t a perfect match for my adgenda — no robots — but it’s probably about as close as can be hoped for, and I think it’s probably close enough that I’ll be able to use it to pursue some of the ideas I plan to pursue on robots later in my dissertation work in abstract toy environments first.

I did spend the afternoon fighting with Microsoft Robotics Developer Studio, but then I came home to my sweet wife and a big wrapped present in the living room. We went out for a yummy Hibachi dinner, came home, opened presents, watched the last West Wing episode (since Thanksgiving we’ve been burning through the complete seven seasons on DVD, and we watched the last episode tonight. Best. Show. Evar. I’m sad there’s no more.), and ate delicious orange chocolate chip cake.

And my Facebook wall has exploded with birthday wishes. Thanks, everybody. =)

Airplane Boarding and Space-Time Geometry

Monday, March 24th, 2008

I’m pretty bummed this talk is happening on a day I don’t plan to be up on campus:

Airplane Boarding and Space-Time Geometry

Eitan Bachmat, Department of Computer Science at Ben-Gurion University (visiting Brandeis University)

It is hard to think of a process that is more boring than boarding an airplane. In the hope of relieving, or at least shortening, some of the pain, airlines have devised various boarding strategies which lead to announcements like “All passenger from rows 40 and above are now welcome to board the plane, or more recently, “Zone 1+2 passengers are now welcome to board”, or even unassigned boarding. In the talk we will try to overturn the negative image that airplane boarding has and will try to portray it as a very exciting process. We will present an analytical model for airplane boarding and try to figure out what are the better strategies. Surprisingly, the model is based on the same math that models relativity theory. In particular, if time permits, we will use airplane boarding to suggest a nice interpretation for Einstein’s law of motion.

I have a thesis topic!

Thursday, February 21st, 2008

Or at least, the germ of a thesis topic. Monday and Tuesday I had good meetings with my co-advisor and advisor, in which I told them my latest thinking and plans for what I want to do for my dissertation. They were both very positive; my advisor even uttered the phrase “I think you’ve really had a breakthrough”. For months I’ve been going back and forth with my advisor, iteration after iteration refining ideas. He kept telling me just “this needs to be more concrete” — I thought I needed to have more ideas, but it turns out he just wanted stick figure illustrations of robots. I still have some details to work out, to say nothing of all the actual empirical engineering work to do and writing before I have a proposal, of course, but at least I can now say I have a direction. It’s about f-ing time. If I’d known all it was going to take was a page of stick figure robots, I would have done that months ago. ;-)

For those of you who are wondering, that basic direction will be to program a robot to exhibit an autonomous developmental progression similar to a human infant, learning to see, scoot, manipulate, stand up, and locomote, with development structured through maturational and motivational mechanisms.

post-ICML-submission mini-weekend

Sunday, February 11th, 2007

Well, I got off a submission to ICML, though the experiments section didn’t have nearly as much in it as I had hoped. First feedback comes in a month, and I’m not terribly optimistic, but I’m glad I submitted anyway. The deadline was 11:59pm in Samoa, hence 6am on the east coast, and I stayed up ’til the last minute in the hopes that experiments would finish that I could put in the paper. (They didn’t.) Strangely, after this whole painful process I’m feeling more envigorated about this work than I have in a while. Can’t explain that.

After sleeping it off, I’m in NYC for a much needed, albeit unfortunately abbreviated, weekend. Tomorrow it’s back to the grindstone, but hopefully with a bit more sleep and a bit more cheer. Soon, I hope: a blog entry that isn’t about work.

New rules for research coding

Saturday, February 10th, 2007
  1. Use a statically compiled and type-safe language. No more running shit for days only to discover that you forgot to include a file or mistyped a method name.
  2. Use a faster language (compilation will help). Python isn’t cutting it.
  3. Cache intermediate results — persistence. If your experiment generates a certain kind of data and then does something with it, write the data to disk so it can be reused. Saves redundant computation, and gets some useful result out of crashed programs.
  4. Learn how to do unit testing, and test the hell out of everything, including independent analyses of scalability.
  5. Visualize everything.  Every module should allow visualization of its behavior and output, for easier debugging and analysis.

There should probably be others, too. Suggestions, fellow computer scientists (or others)?