I try not to let hype steer my Bayesian priors on these things, but my actual *personal* belief is RH may very likely fall by the end of next year.
Sure, people want to see F_1-geo, noncom algebra, Deninger cohomology, etc, but I think AI will soon be capable of theory building.