On Scott Adams' Blog: Robots Program People:
It won’t be long before all new drugs are discovered by robots. This start-up is an example of that trend.And it won’t be long before IBM’s Watson can diagnose and prescribe treatments better than any human doctor.Put those two trends together and robots will be programming humans with drugs. Drugs are the user interface to our moistware.[...]Someday, for sure, machines will be programming humans. And that day will probably be in your lifetime. But don’t be afraid because the robots will someday have a drug that will make you feel totally okay with being their pet.
I think this is quite relevant as that is actually quite likely and has lots of failure modes from UFAI to Moloch.
The post also has some discussion of whether the machines are actually actors in this ("You are thinking that humans program the machines, and the machines just do their job.") and does some insightful comparisons on that.
There's a lot of handwave here.
Meh. Drugs are by definitions substances injected in the body, and cannot discriminate between structurally identical but functionally different neurons.
Unless by drugs you intend something that is totally not like a drug, but then it becomes a standard UFAI scenario.
Who says that the drugging Watsons are the only working component in the system? I think the far-reaching effect of this not-that-unlikely szenario has the potential to make many other failure modes much more likely. A self-satisfied society has much less energy to detect them for example. A more complex system makes failure modes more numerous.