MLUX? The User Experience of Working with Machine Learning

Presented at MeasureCamp. January 2021

Transcript

1. MLUX? The User Experience of Machine Learning
2. MLUX? Lots of different angles to approach this from: • Consumer recommendations: Netflix, Spotify, Youtube etc. • Programming frameworks: scikit-learn, tensorflow, tidyverse • Analysts: Google Analytics Alerts, someone else’s colab notebook, search query reports, data driven attribution
3. Why am I interested in this?
4. One end of the scale • Not many people can do this • Not possible to do this for many tasks
5. Other end of the scale • No control • Only visible after the fact
6. Other end of the scale Graham Everitt (@allcorknobottle) writing on ecommerce guide https://ecommerceguide.com/news/how-googles-move-to-automation-costs-my-company-500k-a-year/
7. Bifurcating into two worlds? I hope not!
8. The way forward? 1. Transfer Learning https://www.eanalytica.com/blog/robocox/
9. The way forward? 2. Control over more inputs
10. The way forward? 3. “Interpretable” Machine Learning https://conductrics.com/machine-learning-and-human-interpretability/