Setting aside all considerations of Trump and his Russian entanglements for a moment, it's interesting how machine learning mirrors our own preconceptions back at us. It can be nice and validating, but not necessarily insightful when applied to public opinion. All it means is that Trump being subordinate to Russian interest is the common expressed perception. That doesn't make it wrong (at all, in this case), but we didn't need an LMM to tell us that.
It's far more useful when applied to seeking out trends or patterns in scientific datasets where those are considerably less apparent. We really shouldn't use this technology to build echo chambers for ourselves.
It's hardly surprising. He's always wanted to be Tony Stark, e.g. simultaneously a nerd and one of the cool kids. Unfortunately for Musk, Stark could pull that off because A) He's neither real nor a realistic character and B) was a bona fide world-class genius. Musk - who's so far from being a genius that you need the JWST to resolve 'smart' from his location in intellectual phase space - characteristically just managed to concurrently fail at both.