ML Safety Scholars Summer 2022 Retrospective — LessWrong