a personal project that i've come to appreciate is my screenshot garden, a small curated mirror of my desktop screenshots folder.
it is really just an online directory tree with a few automated processes on top, but even so i've liked having it- a little folder of time spent browsing, working, and collecting which bundles itself up for you
i've written out some instructions and template code for the workflow, so you can build your own;
In Brazil's main newspaper, Folha de S.Paulo, I argue that microblogging itself and National Novel Generation Month as initiated by @darius are more important to computer-generated literature than particular technologies like GPT-2 and -3 https://www1.folha.uol.com.br/ilustrada/2021/01/robos-ja-escrevem-de-poemas-e-microcontos-no-twitter-a-romances-inteiros.shtml
Two things I like in the "Getting Unstuck" ebook sampler I just released:
* "Who this book is for and what you should get out of it" but also "Who this book is NOT for"
* concrete examples and exercises to improve open source project management skills
I made a toy that generates "What vibes do I give off?" style memes from Wikipedia categories. Enjoy.
conditional dcgan progress
I sorta gave up on having the same model produce different fonts—it just didn't work and the samples across classes weren't similar for the same latent variable (which was the effect I was going for in the first place). HOWEVER, I am super pleased with the samples from the model I'm training on Garamond italics...
i made a mashup EP! it’s called Jaw Bra--20 minutes of high-energy throwback bops to get you thru winter lockdown ❄️💖
arranged in Fuser, lightly edited in Audacity.
hope you enjoy! xoxo
conditional dcgan progress
this is so tantalizingly close to what I want—I'm training the GAN on images of words, conditioned on labels for different text styles (italics, all caps, title case, etc)—you can clearly see many of the different styles in this sample (trained on about 100k images). I managed to avoid mode collapse, but the GAN unfortunately fails to converge (after 200k images, the generator just makes white noise)
logic - pitch us!
we're now accepting pitches for our next issue of Logic! the theme is DISTRIBUTION.
more on the theme, rates, how to pitch, and deadlines are here: https://logicmag.io/pitch-us/
outside of reported pieces, we're also always looking for interesting folks to talk to for interviews (identified or anonymous). our anonymous series talks with rank-and-file tech workers about their work.
if you know of someone who'd like to chat, contact us the same way as a pitch or DM me for my signal
what do you call a gan that doesn't work
Folks with experience in nonprofit development and marketing: https://pyfound.blogspot.com/2020/12/psf-is-hiring-director-of-resource.html The Python Software Foundation is hiring. Remote full-time job; I can recommend the Executive Director and other staff (I have contracted with them frequently).
"doing this" = using DistilBERT to gradually transform a sequence words picked at random from a word list into text that appears to make sense
okay I THINK I finally found a way of doing this that comes close to meeting all of my criteria for this project (i.e., each step shows visible and meaningful change; the change is gradual, but the result "converges" after relatively few steps): calculate the probability of token in source text vs. token sampled from the distribution of mask token at that position, then find "peaks" of improbable tokens, and replace w/sampled token at those peaks; stop when any output repeats
I gave a set of 2020 headlines to the neural net GPT-3, whose training data cut off in October 2019.
Having never seen 2020, it tried to predict what headlines are next on the list.
Poet, programmer, game designer, computational creativity researcher. Assistant Arts Professor at NYU ITP. she/her
Hometown is adapted from Mastodon, a decentralized social network with no ads, no corporate surveillance, and ethical design.