Evolution, bias and global risk — LessWrong