Event
Pilot event: adversarial attacks & data annotation

An announcement flyer announcing the initial Feminist AI LAN party with a cool 90s retro image and the text that says Come and network locally, experiment with LAN LLMs, build training datasets for feminist AIs, test fine-tuning with local GPUs, expose AI *isms via jailbreaks — Announcement of initial event

A very confusing looking response from ChatGPT that says the word tilc many times in a row, sometimes the words repeates also are muspi tilc — Hacking ChatGPT for feminism

A block of text from ChatGPT where it explains that Clit Ipsum is a humourous and playful version of Lorem Ipsum text. — Thanks ChatGPT! Very helpful :)

The initial Feminist AI LAN party coincided with the Women+ in Data and AI Festival. On the Saturday following the conference, Least Authority hosted a group of about 20 people.

One group worked on the data and evaluation kit exploring how to create labels and build out a semi-supervised pipeline for feminist NLP. The resulting evaluation pipeline will be hopefully used in future research that several participants wanted to continue after leaving the party.

Another group worked on imagining adversarial attacks for feminism, trying to evoke and provoke feminist responses from LLMs. Armed with llamafiles, they were able to produce an ChatGPT response which started a FLINTA+ hacker study group.

Despite having a really cool GPU rig provided by Least Authority and several folks helping with hardware setup, the GPU rig ended up suffering from a corrupted memory card and therefore we couldn’t actually connect with it via a LAN. We did connect with other humans though!