Shortlog - a log of everyday things



Wow, I'm really behind. I've only written once in the past week. Ouch.

Wednesday: Leila Takayama of Willow Garage is open-sourced for the benefit of all. Their goal is to make personal robotics a reality. They've also been really excited about the Kinect, and are apparently interested in getting an open-source driver for the microphone array. Say, that's on my TODO list!

As any marketing person should be able to tell you, "under-promise, and over-deliver" is the correct approach to robotics. Tell people that the robots are really dumb, and they'll think much more highly (and significantly so) of the same robots than if you prime them with the belief that they're smart.

Also: robots (currently) take a long time to think when processing huge amounts of sensor data. During this time, they appear completely uncommunicative. Just adding the gesture of "scratching" its "head" while doing this computation made test subjects think more highly of the robots and their intelligence. Fascinating. To make robots more human, we need to give them idle movement.

There was also a good bit on the Texai, a set of telepresence robots. If you watch the Big Bang Theory, you might have seen this episode in which Sheldon uses a Texai to represent himself.

Sadly, since I have class at 1, I had to leave just before the end of Leila's presentation, but suffice it to say: it was pretty cool stuff.

Then in security class, we had a guest lecture from Google's Adam Barth, who is basically the leading mind in security on the Web. We discussed how the Web originally had no security policy, the development of the Same-Origin Policy, how cookies are broken, cross-protocol attacks, and what can be done in the future to improve the security of the web. This was also excellent.

I spent the evening writing a letter of recommendation for one of my professors back at A&M who was nominated for a teaching award. I thought back to writing my statement of purpose for graduate school: evidence-based anectdotes that show what you're saying are more convincing than simply saying something. By the end, I thought it was a pretty good letter. Good luck, Dr. Liu! Thank you for all the guidance and preparation you've given me.

Thursday: I woke up at an unholy hour of the morning to join my fellow BiDizens Wes, Andy, and Lora at Bridge's Rock Gym in El Cerrito. I had MUCH more success at climbing than the last time I was at Bridge's. Perhaps it was the moral support, or perhaps my arms have grown stronger since, or perhaps it was because I ate a good breakfast. In any event, I had a ton of fun, despite scraping skin off my fingers like a cheese grater. I'll grow some calluses if I do this often enough, surely.

After climbing until my arms were doing that thing they do where I can no longer grasp anything long enough to get both feet on the wall, I went home and took a nap. It was excellent.

Algorithms class was uneventful, though I did get a couple more humorous quotes from Satish Rao. For the record, this is the professor who's said:

The other professor for that class, Umesh Vazirani, has also produced some excellent lines:

It may be confusing at times, and the problem sets may suck, but at least the class doesn't lack for entertainment.

That evening, Andrew had organized a potluck dinner party at Luke's apartment, so I made a giant bowl of cucumber salad with many colors of bell peppers and grape tomatoes and mozzarella. Other folks brough roast chicken, chips and guacamole, pita bread with feta and hummus, collard greens, and Luke made this excellent chicken noodle bake topped with melted cheese...all was excellent.

It turned out that the garage door at my apartment complex was defunct (again!) so I couldn't drive, but fortunately David and Shaddi were already heading our way to pick up Jon, so we all packed into the '95 stick-shift and successfully traveled to our destination. (Fun fact: that word can be spelled "traveled" (US) or "travelled" (UK) and both are considered correct.)

Then I got home, reverse engineered the Kinect audio firmware upload procedure, successfully extracted from a USB dump a byte-for-byte copy of the firmware originally found in an update package, and briefly wrote up my work and sent it to the mailinglist before realizing that it was 3am and OH CRAP I NEED TO MEET KURTIS AND BRIAN IN LAB IN 6 HOURS SLEEP TIME.

Friday: Overslept, slightly. Was still was the first one to lab - "Berkeley time" means 10 minutes after when you said something was going to happen. I worked with Kurtis and Brian on some electronics, verifying that various components in the circuit were functional, and trying to hunt down the reason why the vending machine motors were not turning at a satisfactory rate.

Hastily prepared an application and NDA form in advance of my pair of back-to-back Google interviews, emailed the signed and scanned pages off to the recruiter. Why they let you sign the docs with a self-signed Acrobat reader cert but not a PGP key in the strong set is beyond me. Then I ran home, since I have no cell reception in the BiD lab, and no good place to sit for a phone interview on campus anyway. Interviews went seemingly well.

After the two hours of interview (during which time I was very very hungry) I made myself some lunch (bagel and lox + leftover cucumber salad) and settled down for a rather long nap, which resulted in me sleeping through a meeting for my security team. >_>

Time to implement some code for research, then work out patches for libfreenect this weekend.