I have a question about AI Security Research that may be useful for other people working with it too.
I am particularly interested in future improvements and limitations. I would like to compare approaches that people have actually used rather than just theoretical suggestions.
What trade-offs...
