The question is not whether superintelligence is possible, or whether recursive self-improvement can get us there. The question is whether widespread automation will have already transformed the world before the first superintelligence. See point 4.