abramdemski comments on Robustness as a Path to AI Alignment

[abramdemski]

Looks relevant at first glance! Untrusted data is one of the cases I wanted to think more about.