niplav comments on mattmacdermott’s Shortform

[niplav]

Interpretability might then refer to creating architectures/​activation functions that are easier to be interpreted.

[mattmacdermott]

Yep, exactly.