1. Exquisite Tweets from @vtcraghead

    blechCollected by blech

    I did some doomsaying last night re: @justinobeirne's meticulous reverse-engineering of Google's geodata infrastructure: twitter.com/vtcraghead/sta…

    And it needs following-up, wrapped around the idea of "imagination".

    A thread:

    Map/ @justinobeirne.com/google-maps-mo…

    Apple can't keep up, even with its war chest.

    OSM sure as s*** can't keep up, hobbled by gatekeepers who don't want it to.

    Reply Retweet Like

    Google has been the mapping incumbent for a decade. They've surprised everyone before with their ability to build a smart, disruptive (ugh) map hack and then capitalize on it.

    Reply Retweet Like

    But the sort of long-view thinking Justin exposes here is way beyond what I (a lowly map-saltminer) was imagining. Here's what I honestly thought Google was up to over the past ten years to keep their edge:

    Reply Retweet Like

    1. Hire thousands of analysts, to pore over and digitize . . .
    2. A big pile of satellite and aerial imagery, while
    3. Manually ground-truthing addresses with these nifty Streetview cameras, and
    4. Hiring thousands of additional analysts/salesfolk to manage a large DB of POIs.

    Reply Retweet Like

    Bonus, "Wow, they're on the cutting edge", imagined-Google-activities:
    5. Build custom tools for the 1000s of analysts ("Looks like JOSM!")
    6. Buy satellites instead of imagery ("So forward-thinking!")
    7. Experiment with auto-extracting features ("We've been trying since 2002!")

    Reply Retweet Like

    I assumed these things because this is what everyone else has been up to, to one degree or another. Mapping the world is a cumbersome, resource-intensive, largely-manual process. Whew, I'm tired and my hand hurts from all the digitizing.

    Reply Retweet Like

    Reader, my imagination was puny.

    What Justin makes clear is that Google has vertically-integrated their ubiquitous technology in ways that seem handed down from Carto-Olympus.

    Reply Retweet Like

    Google bought satellites (& probably planes & helos), sure, but the buried lede is that they've NAILED DOWN the machine intelligence needed for feature extraction. Their long experience in ML, parallelization, & data organization likely let them brute-force this hard problem.

    Reply Retweet Like

    Sidebar: This is a HARD goddamned problem - one that I've worked on - and so this point impresses the hell out of me (and makes me jealous, but hey).

    Reply Retweet Like

    Google also hired lots of analysts, sure, but they're not likely to be the tip of the mapping spear. Instead, that would be you and me, training a collection of algorithms with our phones in a mad swirl of telemetry and aggro crowdsourcing.

    Reply Retweet Like

    Google has a huge team grooming their "Places" dataset, sure, but those entries aren't all recorded by someone sitting next to a phone, waiting for Papa Luciano to call in with the location of his pizzeria . . .

    Reply Retweet Like

    And most importantly, they're not doing all this in a vacuum, and it's not just for maintenance. Google is integrating these datasets to automate the creation of other datasets, and god knows what else.

    Reply Retweet Like

    They're doing these things in a properly-imagined effort to map our world at the highest spatial, temporal, and social resolutions we've ever dreamed of profit. (Contact your local Borges or Saramagó enthusiast today)

    Reply Retweet Like

    And at this point I need to take a step back and say again that this is all sort of awe-inspiring. I am amazed at the grandeur of it all. I'm envious, because my own imagination was only nibbling around the edges of this holistic approach.

    Reply Retweet Like

    I'm also frightened. This is a lot of power, and it's concentrated by a company that has built some of the most restrictive walls in the business around its data and services. twitter.com/saikofish/stat…

    I'm frightened, but in a bleak-comedy kind of way, when I consider how far behind the competition is, and how little they collaborate. Apple? HERE? Uber? How many smart folks are doing the carto-labor of Sisyphus in back rooms right now?

    Reply Retweet Like

    I'm frightened when I consider that my own windmill-tilt-of-a-project - @openstreetmap - is eating itself alive instead of getting ahead of this tidal wave.

    Reply Retweet Like

    The Google context makes the nonsense politics of OSM . . .

    - Driving away newcomers and force multipliers
    - Alienating whole demographics
    - Dismissing any technology more elaborate than a GPS and a bicycle

    . . . frustratingly insignificant but completely defeating.

    Reply Retweet Like

    I've seen glimmers of Google's approach in the OSM world . . .

    - Feature extraction by the @developmentseed crew
    - @Mapbox fusing sources for QA/QC
    - @Facebook literally donating data

    . . . shot down by an unbendable - if surficially-admirable - John Henry mentality.

    Reply Retweet Like