> To build the dataset, we keep only sketches that have been recognized by the neural network in the QuickDraw game. To prevent imbalance from overly represented countries (e.g., the US), we down-sample the data by focusing on the 100 most prominent countries and capping the number of drawings per category–country pair at 10,000. [p19]
The most striking real world example of something similar has always been the different ways different cultures count/show numbers on a single hand (Ask a friend to show the number 3 on a hand). As far as concepts - it’s a difference in how we perceive the “starting point” of a hand.
There's a story in Rhodes' book on the atom bomb of Otto Frisch & Liese Mettner discussing the ideation of nuclear fission analogised to cell fission, drawn as a dumbbell viewed head on with the neck of the split a circle inside the bigger cell circle:, where we classically see two cells splitting side by side with a channel between them: she meant exactly the same thing viewed 90° rotated.
The phone cluster was especially amusing to see in the context of visiting a small history museum in Ohio earlier today with my kids (12 years old) and explaining to them about how a rotary dial phone worked.
(The added bonus comes from watching Monsters and Minions with them yesterday and during a scene where the director is informed that they ran out of film, my daughter turned to me and asked, “What’s film?”)
Sometimes this is exaggerated. How do people view slide rules, telegraphs, and gramophones, or other historical artifacts from a further generation ago? They aren't that strange, and neither should rotary phone or film be to the current generation.
I agree on your general idea but rotary phones were a real PITA.
They were slow to operate. We jumped in a very short time to numberpad phones, which initially only replayed the analog signal of rotary phones. I'm surprised that nobody built an analog phone with a numberpad that operated a hidden rotary dial before electronics got everywhere. It would probably need to be plugged into main but those phones were not mobile so a socket was always available.
On the other side, a 33 or 45 rpm record is only inconvenient to store and transport but it plays back in the same time of any modern music format and seeking is not so bad: you can see where you move the head and with a little of practice one gets quite good at it.
The slowness of rotary phones is coupled to the exchange, which expects pulses at a specific rate (about 10 per second). Shifting to a keypad form factor doesn't speed that up, and would mean you need to be careful about how quickly you dial (or have a way to buffer keypresses). I think that's a major reason why keypad dialing and DTMF signaling were so closely connected.
I know how it worked. Buffering (probably complicated on a pure mechanical device) or locking the keys until the dial came back to rest would still be better than rotating and misrotating the dial.
https://github.com/googlecreativelab/quickdraw-dataset
reply