The GradientJul 11, 2023
Naomi Saphra

Super pumped that @thegradient has published "Interpretability Creationism", my manifesto for considering the training process in AI interpretability research!

https://thegradient.pub/interpretability-creationism

Interpretability Creationism

On “interpretability creationism” – interpretability methods that only look at the final state of the model and ignore its evolution over the course of training

The Gradient
Jul 11, 2023 at 7:37pmWeb
Show threadDavid RuffnerJul 12, 2023

@nsaphra @thegradient

I really like the idea of observing what happens throughout training. I've done some of that when training object detectors and it was helpful. And it is fun too! It is really cool to see how the system learns.

Show threadA.V.Jul 12, 2023
@nsaphra well written.