A large portion of generative AI is based on Bayesian statistics, like stable diffusion, regularization, LLM as a learned prior (though trained with frequentist MLE), variational autoencoders etc. Chain-of-thought and self-consistency can be viewed as Bayesian as well.