Second, I have got the feeling that many people who experiment with genetic manipulation don't know what they are actually doing. Yes, they might know it on a high-level perspective and for experiments that is fine, but as long as we can't simulate even single cells its is hard to tell that we know what we are doing.
As a computer scientist it feels to me like someone is patching the binaries of 'grep' and 'cp' together and hopes to use the resulting binaries to build autonomous drones. It is just irresponsible to use the results are long as the have only one petri dish which is actually the one we are living in.
So don't get me wrong, I am completely pro research, but also completely against usage outside of laboratories. There might be exceptions to that rule when it comes to 'specific modification' as you call it, but for me it ends where you could not achieve the same result with a handful of normal breeding cycles.