Instructor: Anastasia Davydova Lewis | xnastasia on Telegram Location: Situated Intelligence | Online Time: April 20, 2021 12PM PDT Tools: bash, smartphone, computer with internet, RunwayML,Scrapism, or nothing.
Most approaches to teaching machine learning focus on specific models and methodologies. While we are constantly discovering new techniques, one thing remains consistent: the quality of a model’s predictions is always going to depend on the quality of its dataset.
In this workshop, we will design and curate a dataset, and preview what an ML model that you can build yourself would look like.
introductions
overview: mini lecture slide deck
- permanence of bias challenge
- what is bias
- what is scrapism
- what is escrapism why like escapism haha
- how to be honest in avoidant practices
- acceptance of inherent bias in data representation / data vizualization for machines
7 types of Data Bias in Machine Learning
bash tutorial sam lavigne's command line intro
- make a folder
- navigation
- now we need stuff to put in the folder
- What sort of data do you have? Notes, photos, videos, receipts, spreadsheets
- Find data vs Create data
- Label it
manipulating data and labels resizing renaming
- Runway models
- stylegans
- text processing
- utility of different models
more bash
- make some nice representation of the dataset
hanging out
- Hyperdrive Portal from HDSA Becoming a Server
- Training and the Problem of Data
- Daniel Shiffman's Introduction to Runway: Machine Learning For Creators
- Gene Kogan's Machine Learning for Artists
- MIT Unbiased Look at Dataset Bias
From Sam Lavigne's class:
- The Cut Up Method by William Burroughs
- Uncreative Writing by Kenneth Goldsmith
- Digital Divide by Claire Bishop
- Montage by Jared Leibowich