I have only just read the mesa optimizers paper, and I don't understand what it adds to the pre-existing picture that "ML can fail to generalize outside the train distribution and this is bad."

The discussion in the paper generally assumes a background distinction between "training" and "deployment" (or an equivalent distinction), and discuss models which succeed on the base objective during "training" but not during "deployment."

In the sections about "deception," this happens in a special way quite unlike the ordinary failures to generalize that we see in ML today (and arises under correspondingly exotic conditions). But, in cases other than "deception," the paper describes dynamics and outcomes that seem identical to the ordinary generalization problem in ML:

Training find a model that scores well on the base objective, assessed over the training distribution
But, this model may not score well on the base objective, assessed over other distributions

For example, the following is just a textbook generalization failure, which can happen with or without a mesa-optimizer:

For a toy example of what pseudo-alignment might look like, consider an RL agent trained on a maze navigation task where all the doors during training happen to be red. Let the base objective (reward function) be $O_{b a s e}$ = (1 if reached a door, 0 otherwise). On the training distribution, this objective is equivalent to $O_{a l t}$ = (1 if reached something red, 0 otherwise). Consider what would happen if an agent, trained to high performance on $O_{b a s e}$ on this task, were put in an environment where the doors are instead blue, and with some red objects that are not doors. It might generalize on $O_{b a s e}$ , reliably navigating to the blue door in each maze (robust alignment). But it might also generalize on $O_{a l t}$ instead of $O_{b a s e}$ , reliably navigating each maze to reach red objects (pseudo-alignment).

Additionally, when the paper makes remarks that seem to be addressing my question, I find these remarks confused (or perhaps just confusing).

For instance, in this remark

The possibility of mesa-optimizers has important implications for the safety of advanced machine learning systems. When a base optimizer generates a mesa-optimizer, safety properties of the base optimizer may not transfer to the mesa-optimizer.

I don't understand what "...

Academic Papers