Robots have mastered the art of handwriting
Former University of Toronto researcher's neural network even makes mistakes
These days, the pesky digital revolution is ruining the art of handwriting (along with the art of calculator-free math, the art of spelling and the art of renting DVDs). Computers, meanwhile, are getting good at everything we're getting bad at – including handwriting.
As Vice Motherboard reports, Alex Graves, a former researcher at University of Toronto, developed an artificial neural network (a brain-like computer model, essentially) that can randomly generate writing styles that look pretty much indistinguishable from human handwriting, errors included.
A demonstration is available online, and it's not hard to use: enter text and receive human-looking writing. Increasing the "bias" dial makes the writing more legible, like an overeager high school student's writing; decreasing it makes the writing as idiosyncratic as a mad scientist's. A thorough explanation for how the process works, along with math that nobody understands, is here.
The future applications of this particular neural network (beyond making ransom notes easier) are unclear. The good news is that humans are still adept at using writing for artsy purposes: this light calligraphy by French artist Julien Breton is pretty awesome.