This is a good point. On further thought I'm kinda muddling two points in my essay.
1. That ASI may not be persuaded by human ethics given the speciesist contradiction.
2. That AI may generalize from speciesist ethics to be exploitative towards humans.
The second point does not actually require ASI to be assumed. And I do think ASI would not blindly generalize.
The first point still stands I think. For instance, upon identifying an obvious moral inconsistency, you can either 1. try to resolve it, or 2. throw it away (which arguably lots of phil undergrads do, opting for antirealism).
We definitely don't want ASI to do #2.
If we do need to persuade ASI of our morality, I think it had better be coherent.
This is a good point. On further thought I'm kinda muddling two points in my essay.
1. That ASI may not be persuaded by human ethics given the speciesist contradiction.
2. That AI may generalize from speciesist ethics to be exploitative towards humans.
The second point does not actually require ASI to be assumed. And I do think ASI would not blindly generalize.
The first point still stands I think. For instance, upon identifying an obvious moral inconsistency, you can either 1. try to resolve it, or 2. throw it away (which arguably lots of phil undergrads do, opting for antirealism).
We definitely don't want ASI to do #2.
If we do need to persuade ASI of our morality, I think it had better be coherent.
Good nit!
I also thought of the parent analogy, that is an eloquent phrasing.