hi fediverse, I just posted transcripts of a few talks and lectures I've given over the past few years, mostly concerning the connections between machine learning, language and poetry: https://posts.decontextualize.com
(notes and summaries in individual posts below)
I'll be part of these free, outdoor, computer-generated performances in NYC, Thu 7pm, along with @aparrish & others https://hbstudio.org/calendar/performing-algorithms/
alright yeah shout out to @jonbro for this idea. here's the Deep Space Nine introduction made entirely using a Sketchfab browser preview of a 3d model of deep space nine (with original musical accompaniment)
one thing I hate about twitter is having zero control of your posts’ reach. it strips agency, flattens context, contributes to uncontrollable harassment in the service of “connection”. imo mastodon’s insistence on following the same value of maximization, removing any ability to foster small local community, is equally harmful. if decentralization is about control then why strip that at the individual and community level? it’s no better than twitter. glad to be on the hometown fork
HomeTown is a fork of Mastodon which lets you make instance-only posts, adjust character limits and read long form rich text blog posts.
It's still part of the Fediverse, and HomeTown users can interact with Mastodon users totally fine.
Tech people can get self-hosting instructions here:
Non-tech people can use a managed hosting service to start their own HomeTown instance:
a few weeks ago I made a tutorial for how to use the Hugging Face Transformers python library to generate text, focusing on the distilgpt2 model in particular. the tutorial explains subword tokenization and shows a number of simple (but effective) ways you can control the model's output: https://github.com/aparrish/rwet/blob/master/transformers-playground.ipynb
computer-generated "recipes" that I made as an example in the workshop I'm teaching. the instructions are composed of random transitive verbs plus random direct objects from Bob Brown's _The Complete Book of Cheese_ https://www.gutenberg.org/ebooks/14293
doesn't do so well at the inverse task, i.e., generating with the probabilities of any token containing a vowel letter OTHER than 'E' zeroed out
#Github #Copilot gives an idea why #Microsoft paid so much for Github. They were after data: Tons of food for their AI, millions of contributors that now 'work' for MS for free.
You publish your code under GPLv3, even AGPLv3? So what? The AI learns from your code and uses it to generate code that is possibly proprietary. Does #GPL forbid this practice? (I don't think so)
That's the M$ way to break copyright law.
It's time for alternatives like @codeberg .
love this sorta disgusting visualization of a self-organizing map https://www.complexity-explorables.org/explorables/yo-kohonen/
Lately I've been reading a lot of children's picture books, over and over? I thought "Goodnight Moon" was pretty spooky, but I had trouble finding anyone writing about that online. @redoak jokingly suggested that I become the conspiracy theorist blogger I want to see in the world, so... I did it. Here's a totally serious take on why "Goodnight Moon" is an esoteric text, from me, a serious scholar of esotericism (aka podcast listener): https://pseudony.ms/blags/goodnight-nobody.html
logit biasing, markov chain style. here I'm doing it with phonetics—basically I check the possible outcomes for each context, and then artificially boost the probability of predictions that have certain phonetic characteristics. (in this case, more /k/ and /b/ sounds)
(tomorrow I'm going to see if stealing alternatives from similar ngrams helps... but I am beginning to more viscerally understand why the solution to language modeling that really caught on is just... More Training Data)
I like having this extra setting to fiddle with! but based on my limited testing, the temperature doesn't really matter once the length of the ngram hits a certain limit, since most ngrams only have one or two possible continuations. like... with word 3-grams, it's pretty difficult to distinguish 0.35 from 2.5
generating with a markov chain using softmax sampling w/temperature (a la neural networks). this is an order 3 character model, and you can really see the difference between low temperature (instantly starts repeating itself) and high temperature (draws from wacky corners of the distribution) (if you've generated text with a markov chain before, it's probably using what amounts to a temperature of 1.0)
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.