It Is Fucking Cold Outside

As I stood in the record setting cold this morning waiting for public transportation, I struggled in a way that I’m sure many other of us in this city, on this cold morning struggled as well. As I kept my hands exposed to play games on my phone, they got colder and colder. And eventually I had to make that Sophian choice between boredom or frostbite. I chose boredom, electing to preserve the functionality of my hands and opposable thumbs for future mobile gaming opportunities.

I pray I made the right choice.

– Someone inserted Will Smith into the grand Star Wars mythos in this delightful Independence Day: A Star Wars Story mashup, and I am eternally grateful. I’ll never know why the universe took two decades to put this in front of me, but it doesn’t matter. I have realized my destiny.

– Much has been said here already about the merits The Big Bang Theory, so I don’t really want to ramble on again about how funny or unfunny the show is, I’m just here to make one last point about it. We’ve all pointed to these awkward scenes with the laugh track removed as “evidence” of how unfunny the show is, never mind the fact that removing a keep component of a comedy show’s format causing it’s pace and timing to be disrupted and make it feel kind of weird is not something that is necessarily unique to this show in particular. The laugh track serves the format of a multiple camera sitcom. I mean yes, the actual lines they’re ending on to elicit the audience laughter less punchlines and more…just general statements, but would Seinfeld still be funny without a laugh track? …actually yes, but okay, fine whatever. Anyway, my point is…replacing The Big Bang Theory‘s audience laugh track (or the legitimate sounds of oddly enthusiastic laughter arising from being “filmed in front of a live studio audience”) with the prerecorded sound of Ricky Gervais laughing raises it to a totally new height, transcending the sum of its parts. I would watch 18 to 22 minutes of this every week if it was available. I mean…some of it becomes like laughing at “him” because you’re wondering what he’s picking up on that’s so funny…and then the whole thing becomes even funnier. Layers upon delicious layers.

– Here’s an interesting article in which some dudes (researches, scientists, programmers?) are using deep learning and Google Street View to estimate the demographic makeup of neighborhoods across the United States. Pretty nifty correlations are seen when using 50 million images of Google Street View to identify objective visual details across neighbourhoods in 200 American cities. Cars in particularly seem to be the most useful to look at since “over 90% of American households own a motor vehicle, and their choice of automobile is influenced by disparate demographic factors including household needs, personal preferences, and economic wherewithal”…so you can tell a lot from a neighbourhood’s choice of car. As opposed to say…shrubbery or landscaping, spaces between houses, lot sizes, house exteriors, or blurry images of people on the street. So plug all the images into a machine learning algorithm, teach it how to identify makes and models of cars, reference it census and election voting data, and boom goes the dynamite. It’s a neat read, so you should go through it yourself, but some interesting details…

• The strongest indicator that a neighbourhood was primarily Asian was the presence of Toyotas. Secondary indicators include…minivans, Hondas, and cars made in 1990-1994. Using all factors together their model was able to identify demographic attribute of “Asian” pretty accurately (r=0.87).
• Strongest indicators for white neighbourhoods were Aston Martins, Volkswagens, minivans, Jeeps, and extended cabs. Was a bit harder to nail down the whites, as there’s some more scattering here, but still not bad (r=0.77).
• They get better at predicting black neighbourhoods (r=0.81) which have top indicators of cars made in 2010-2014, sedans, and Chryslers, Buicks, and Oldsmobiles.
• The indicators are actually better at predicting Asians (r=0.87) than at predicting median household income (r=0.82). The latter of which I would have thought would be easier to find based on just presumably a pretty strong correlation with the price of the car…but I guess Asians driving Toyotas are more likely than rich people having expensive cars.
• Cities that have more pickup trucks than sedans were more likely to vote Republican (88% chance). On the flip side, cities that have more sedans than pickup trucks were more likely to vote Democrat (82% chance). It’s that simple, we’re all just numbers in the end.

I know we had stereotypes in our heads already, but now they can be collaborated against objective statistics! And this is all just based on studying 500 million pictures of questionable quality from Google Street View, which included 22 million identified vehicles, representing 32% of vehicles within these 200 cities, or just 8% of total vehicles in the United States. As self-driving cars that use cameras to navigate start to populate the roads – “Tesla vehicles currently take as many images as were studied here every single day” – there’s going to be a considerably larger sample size to study. So if a half dozen guys from Stanford University can put this together with what amounts to just a virtual stroll around American neighbourhoods, you gotta wonder what the genius tech giants out there have already figured out based on much deeper cuts of our lives available to them.

I have contained my rage for as long as possible, but I shall unleash my fury upon you like the crashing of a thousand waves! Begone, vile man! Begone from me! A starter car? This car is a finisher car! A transporter of gods! The golden god! I am untethered, and my rage knows no bounds!