As a roboticist at the beginning of my career working on drones, I decided then and there that I would never make "bombs", a metaphor I used to mean anything that could be weaponized. I realized a lot of the work I was doing was funded by DARPA, and I was very cognizant about my research being used in this way. And like Dr. Kearns suggests, it's not entirely black and white. Would my path planning algorithm be used to more efficiently deliver scientific payloads to the atmosphere, or would it be used to route missiles to maximize casualties? Hard to really say, but I've avoided overtly military applications (even things like BigDog, designed to carry equipment for troops).
Sometimes the distinction is even more insidious. I did work on perpetual flight for drones, and Facebook had a perpetual flight project that had the goal of bringing internet access to remote locations in Africa. Sounds humanitarian, but I also didn't want to be responsible for subjecting poor Africans to what I consider Facebook's panopticon. Or maybe it would have been a net boon for the region? It's really hard to tell a priori, so the best I have managed for myself is just to try and stay in theory, where developments are further removed from direct consequence.
Sometimes the distinction is even more insidious. I did work on perpetual flight for drones, and Facebook had a perpetual flight project that had the goal of bringing internet access to remote locations in Africa. Sounds humanitarian, but I also didn't want to be responsible for subjecting poor Africans to what I consider Facebook's panopticon. Or maybe it would have been a net boon for the region? It's really hard to tell a priori, so the best I have managed for myself is just to try and stay in theory, where developments are further removed from direct consequence.