Question: What role do people who think that AI research is dangerous hold in the field of AI research?
Answer: None...because regardless of their qualifications, they wouldn't further the progress of something they think is a very, very bad idea.
Asking AI experts whether or not they think AI research is a bad idea subjects your responses to a massive selection bias.
Yes. Nobody who worked in the Manhattan Project had any reservations whatsoever about building the atomic bomb, right?
Experts work in fields they're not 100% comfortable with all the time. The actual physicists that worked on the bomb understood exactly what the dangers were. The people looking at it from the outside are the ones coming up with the bogus dangers. You hear things like, "the scientists in the Manhattan project were so irresponsible they thought the first bomb test could ignite the atmosphere, but went ahead with it anyway." No, the scientists working on it thought of that possibility, performed calculations the definitely proved it wasn't anywhere near a possibility and then moved on with it. People outside the field are the ones that go, "The LHC could create a black hole that will destroy us all!" The scientists working on know the Earth is struck with more powerful cosmic rays than the LHC can produce regularly, so there's no danger.
It's just that they don't work in the field of AI, so therefore they must not have any inkling whatsoever as to what they're talking about.
Which is a 100% true statement. They're very smart people, but they don't know what they're talking about in regards to AI research, and are coming up with bogus threats that most AI experts agree aren't actually a possibility.
The topic of the Manhattan Project is a red herring. Those people were choosing between two evils, because the Project was about building a weapon to stop a genocidal maniac from taking over the planet. By the time they were done, D-Day and V-E Day had happened, true, but those victories were far from foregone conclusions when the scientists started.
Nobody's building AI to try and prevent something on the same level as world domination by Hitler, sorry.