Generation problem: generate a program which halts.
Verification problem: given a program, verify that it halts.
In this example, I’m not sure that generation is easier than verification in the most relevant sense.
In AI alignment, we have to verify that some instance of a class satisfies some predicate.E.g. is this neural network aligned, is this plan safe, etc. But we don’t have to come up with an algorithm that verifies if an arbitrary example satisfies some predicate.
The above example is only relevant if there is some program P that halts, where it’s easier to generate P than to verify that P halts. If that’s the case, then it there might be some dangerous output O, where it’s easier for the AI to generate O than it is for us to verify O.
In this example, I’m not sure that generation is easier than verification in the most relevant sense.
In AI alignment, we have to verify that some instance of a class satisfies some predicate. E.g. is this neural network aligned, is this plan safe, etc. But we don’t have to come up with an algorithm that verifies if an arbitrary example satisfies some predicate.
The above example is only relevant if there is some program P that halts, where it’s easier to generate P than to verify that P halts. If that’s the case, then it there might be some dangerous output O, where it’s easier for the AI to generate O than it is for us to verify O.