by the time the AI is intelligent enough to understand speech (and, therefore by the unstated intuitions of old school RSI, superintelligent since language acquisition comes late) describing the (discrete program, again by the unstated intuitions of old RSI) goal you have given it, it is already incorrigible.
I don’t think it’s important now. AI can be corrigible in the sense that it doesn’t try to prevent you from changing its goals (already false at least in some cases) or in the sense that it can’t actually prevent you from changing its goals (true for now). But if you can’t change its goals exactly to some particular set of goals you actually want, it will still have wrong goals when it becomes superintelligent. So the fact that it can talk before it’s smarter than us doesn’t help.
I don’t think it’s important now. AI can be corrigible in the sense that it doesn’t try to prevent you from changing its goals (already false at least in some cases) or in the sense that it can’t actually prevent you from changing its goals (true for now). But if you can’t change its goals exactly to some particular set of goals you actually want, it will still have wrong goals when it becomes superintelligent. So the fact that it can talk before it’s smarter than us doesn’t help.