Rob, okay. I see what you're saying and I agree. OTOH, I think it's natural that we will want to learn what we can about emotions while we're learning what we can about intelligence. It's not even clear the extent to which these things can be isolated from each other.
However, it's possible we could run into a problem even if the machines don't have emotions. Contrast Saberhagen's Berserker Saga with Ellison's short story "I have no mouth and I must scream", e.g. (Predecessors to The Terminator.)
In the Saberhagen's story (which is a fun story even though it's pure christian propaganda), the machines have no emotions. In Ellison's, the machine is experiences hate and revulsion at the humans who created it.
No doubt we ought to consider carefully before we go down certain roads (but someone will anyway). It's all quite a ways off, I'd think. (Like I said, I don't agree with Kurzweil.) Still, it's good to consider the possibilities in advance.
The hurdles should not be, ahem, misunderestimated (hehe).
Turing (about the time of WW II) predicted there would be machines that could play grandmaster level chess within 20 years - turns out it was more like 50 or 60 years. (There's honorable precedent for software engineers' inability to reliably predict level of effort and completion time for projects.)
There's another story (which I have not confirmed) I heard once about how (I think it was) Marvin Minsky when he was a grad student was asked to develop a machine vision program as like a summer project. The prof had apparently thought it wasn't going to be a huge deal. Turns out it's the problem of pattern-matching - huge, huge problem in CS. OTOH, we had thought teaching machines logic was going to be hard. Turns out that one is trivial (well, almost). There is an unspoken part of logic - the translation phase - that is laden with assumptions and ambiguity. (It's primarily for this reason that logic is not trivial.) But the actual performance of logical operations is a trivial thing for a computer to do, even though it's apparently a bear of a problem for most humans. We're good at pattern matching. The machines are good at logic operations.
