Let me rephrase to: "Discussions of AI sometimes end, where they have serious implications regarding real life."

I agree, that would be a bad thing.

Closing your case when confronted with possible results from a chain of argumentation won't make them disappear.

Of course not, but given my values and my estimates of how likely are certain future scenarios, I already came to certain conclusions. For them to change, either the values or the probabilities have to change. I find it unlikely that my values will change as the result of eschatological discussions on the 'net, and the discussions about the probabilities of Skynet FOOMing can be had (and probably should be had) without throwing the police state into the mix.

In general, I don't find talking about very specific scenarios in the presence of large Knightian uncertainty to be terribly useful.