Issue 300
— Insight —
Rethinking talent strategy as AutoML comes of age
Data scientists are hard to hire and unfilled positions abound. Will advances in AutoML change the game?
Participation-washing could be a dangerous ML fad
"If we’re not careful, participatory ML could follow the path of AI ethics and become just another fad that’s used to legitimize injustice."
Since a lot of the logic in machine learning systems is in the model parameters, testing ML is very different than testing traditional software systems. Here's a great overview of testing considerations, strategies and a selection of worthwhile links for further reading. This post walks through an end-to-end example of deploying a ML product using Jupyter, Papermill, Tekton, GitOps and Kubeflow.
— Tools and Techniques —
Effective testing for machine learning systems
Data Science Meets Devops
New course from FastAI! Covers a variety of topics in data ethics and is intended for a broad tech audience. It's free and covers things like disinformation, bias & fairness, tools, privacy & surveillance, and more. Nice collection of machine learning survey papers covering a variety of subjects, including Recommendation Systems, Deep Learning, NLP, Embedding and more. Curated by Eugene Yan.
— Resources —
Applied Data Ethics
ML Survey Papers
There are a lot of ways that visualizations can be unintentionally misleading. This post explores common issues and offers a testing strategy for identifying them. Oftentimes, visualizations are presented as finished projects and don't offer insights into the thinking process behind them. In this post, Moritz Stefaner bucks that trend and walks through his process for developing his latest interactive project, The Language of Science.
— Data Viz —
Surfacing Visualization Mirages
How to Turn 175 Years of Words in Scientific American into an Image
No spam, ever.