khafra comments on How can we ensure that a Friendly AI team will be sane enough?

[khafra]

Or, proper mechanism design for the research team might be able to smooth out those troughs and let you use the highest-EV researchers without danger.