Alignment by induction Assume model_0 is aligned (enough). What affordances can we give model_n to increase the probability that model_n+1 is (more) aligned? Model_n and model_n+1 might be over timesteps of a continual learning system, training steps or successor generations of models.To what extent do these methods also amplify misalignment? If a mistake is made along the chain, can we correct it using previous models? Maybe we can be nice to our models and work with them rather than constantly assuming an adversarial stance towards them.
Induction doesn't apply perfectly here, instead we'd like to know how well/badly conditioned our methods are. This is probably over and above the conditioning of the training process per se.
Condition numbers assume that errors propagate independently. Would be cool to build a coding theory for alignment so that we can plausibly correct random errors.