Riccardo, I'm not sure I understand your core argument. On the one hand, you seem to be arguing in favor of a proposition whereby alternative paths to knowledge (e.g. AI) may be more fruitful than simply building larger colliders; on the other hand you seem to be arguing in favor of utility (e.g. "what practical social benefits will likely emerge from doing X"). We know that AI is only as good as the datasets it's trained on, which implies that simply using AI to elaborate already-abstract models is unlikely to result in meaningful breakthroughs. Meanwhile, even $100 billion is a small sum compared to OECD spending on weapons procurement, and insignificant compared to civilian spending on junk food. Yet it's difficult to argue that ever-more-expensive-and-lethal weapons represent a net public good, and it's very easy to argue that consumption of endless junk is making us all very ill indeed. There's surely an argument to be made that knowledge is of itself often a (diffuse) public good. For example, in consequence of the scientific revolution we no longer rush out to burn witches every time something unfortunate occurs. We as a species have a slightly firmer grasp of reality because, aside from the most ignorant and intellectually stunted among us, we understand that the universe isn't about childish gods, goblins, and ghouls but about impersonal fundamental laws. The benefit from this one shift in mental posture is incalculable.