I woke up at 9:30 and decided to go back to sleep. It was actually a good decision though–I had a sequence of pleasant naps with little fleeting dreams and was able to reach out to the omnipresent warmth of simple, unoptimized reality.
I had a nice breakfast, read some RIAI. Hung out with Marcos for a bit (another one of those nice, philosophical guys who’s introverted enough I don’t get to see him very often). I did a bit more work on RIAI then went to a volleyball meetup (needed a rest from all that basketball). It was pretty nice. I played with some noobs and managed to play decently well. I got to know some of my teammates, added a bit of my usual extroversion and high fives and all that. We moved to a game mixed in with the good players and I still held my own. Plus, when we did something awesome, I would head downtown and hop up and grab the hoop (one hand, nearly had 2 once). I don’t think I’ve ever been able to do that before, I could often touch, but not grab and hang!)
Since I was already halfway there, I went down to the Stadthaus (city hall) and got myself deregistered. That means I don’t have to pay insurance for February :D :D
Went back to Culmann, made myself yet another late lunch, an omelet stuffed with Aldi’s goodies that was quite tasty. By now it was nearly 5 and I got back to work. More RIAI, ended up reading one of the original papers on adversarial examples, it was quite interesting; it argued that the existence of these examples that fool neural networks and other machine learning models was an unavoidable consequence of the essential linearity of the models. (Technical note: the models are nonlinear and can in theory model any continuous function on R^n. But they need to be nearly linear to be trainable. So it might be the case that with radically different training methods and nonlinear optimization, we could avoid the adversarial examples. But they are otherwise baked in pretty far.)
I chowed down on some curry, read a great ribbonfarm article about the representation of the sequence-of-finite-games-making-up-an-infinite-game in literature and how that relates to the hero’s journey and carrier bag theories of fiction.
Then Muriel came over. We sat ourselves down in the Culmann common room lounge and cracked some beers. We had a fun goofy time, Muriel destroying me at a version of liar’s dice with only two dice. (It almost triggered a response like getting beaten at Wheeler Dealer by Henry B. way back and then going home and calculating all the investment payoff probabilities and expected values).
She was in for both my random ideas (trying to make a cocktail with grapefruit, sugar and bourbon, and watching Black Mirror.) That left us in my favorite state, lots of ideas and short on time (she had to catch the midnight train). First we got up and played some foosball against Marcus and Fabian and smashed them. Then Marcus and Muriel teamed up against me and I beat them 10-3 (hehehe I can’t resist the Mike Peters-esque sports score reference here).
Then we watched a couple episodes of Black Mirror. It was really good. We kept pausing to insert little analytical comments and relate short stories :D It felt like only a few minutes before Muriel had to catch her train home–we’ll have to try the grapefruit cocktail next time.
Wrote this log and hit the hay!