For about 6 months now, Google’s been doing a little experiment into a) neural networks, and b) how bad the average human is at drawing, say, a watermelon, in the form of its Quick, Draw game. Available here, the simple challenge tasks players with drawing common objects, in such a way that the company’s computers can correctly identify them. It’s neat to see how accurate it generally is (unless these sophisticated programs somehow can’t detect a perfectly well-drawn watermelon, not that we’re still mad about it), but it’s also allowed Google to collect a huge repository of stick figure drawings of a bunch of different things, which the company is showing off and sharing with people today.
Besides the somewhat creepy reminder that there’s literally nothing Google-adjacent you can do that isn’t probably being recorded and analyzed for something, it’s actually really cool to watch how people respond to prompts of varying complexity, from a simple line, all the way up to things like the Eiffel Tower and “animal migration”. The company has several million drawings on file, and it’s weirdly soothing to click through and see how similar, and how different, people’s approaches to these drawings are. (And in the meantime, you can help the company troubleshoot its data sets, the better to ensure its eventual murderbots can accurately identify an owl or a pillow when they launch their inevitable rampages to purge us all.)