Couldn’t sleep, probably the first time in 10 years I was actually forced to get up after two and a half hours of unanswered prostrate pleas to the sleep god. On the plus side, I was thinking about interesting things the entire time, and was grateful not to be frustrated.
So I got up at 4am or so, headed up to the roof, went downstairs and made myself a bowl of cereal, read lecture notes for Game Theory, started on the RIAI Synthesizer project (ungraded, but interesting), read Brandon’s blog, and then crashed again. This time I slept soundly until my alarm went off some 2.5 hours later.
Up and at ’em! I hustled to make myself a delicious eggy scramble and made it to RIAI just in time. Another insane lecture, this time on probabilistic programming–I’d seen it before a little bit but the sheer expressivity of it blows my mind. This is normal (functional) programming, but you want randomness in your inputs and outputs. You can magically assign distributions to variables and, even crazier, condition the current program state on “observations”, and at any time ask for output distributions. The inference task of representing the distributions around and transforming them with the statements of the program is automatically handled by the language backend (compiler/interpreter), either via exact methods or sampling. I’m not sure how far this goes–how complicated can the distributions be? But the applications appear vast.
Grabbed lunch with Martin, talked about the class. Went to RandAlg lecture. Went to RIAI section, which helpfully clarified the lecture.
Went home, did some admin, took a look at some Putnam problems! Then I went to the gym, ended up in the wrong class, which was
Super Kondi Attack!!
The gym was incredibly full. A guy on a platform was doing and calling moves, and everyone was following to an absolutely pounding beat. The workout was pretty exhausting, sweat was pouring. The struggle primed us to yell at the appropriate times–the whole thing felt like a mad rave.
After the class, straggled back to Culmann for some dinner, getting my free beer on the way and stopping in the CSE lounge to challenge some random people to a game of foosball. Me and my partner Francesca destroyed our opponents.
Dinner was delicious, made myself some burgers. Talked with Alexei and Tom and played some more foosball :D
Went up to work on the Putnam problem, made a number of errors and got a bit frustrated. Drowned my frustration by reading about this badass prof called Sandu Popescu, an quantum theorist who had (also) replied to my email asking for the slides to his talk. (Same as Andrea Fratalocchi from the Schroedinger colloquium I went to, the best people are the most responsive, crazy!) Skimmed some of the fellow’s research–it was quite incredible the number of important results he had and across many different parts of quantum theory.
Then, since I STILL even after being totally sleep-deprived had jet-lag insomnia, I took a nice picturesque walk. I came up with my question for these weirdly responsive top minds: How did you pick/end up in your field, and when? How do I choose to specialize between AI and all the cool stuff in physics (quantum information, condensed matter engineering, biophysics)??
Then I wrote this log and crashed.
Gratitude Addendum
- Didn’t get annoyed at myself for insomnia
- Lunch with Martin was fun–he had a cool intuition about combining random variables under different operations
- Super Kondi was really fun, unique experience!
- Got lots of extra foosball in
- Burgers were amazing, came out perfectly cooked
- Probabilistic programming!!
Todo
- Remember to apply to the Perimeter Institute summer program in January (plus their master’s program is cool)
- Ask Popescu and Fratalocchi my question
- Hang out with Alexei more