The reinforcement learning algorithm of facebook, instagram, youtube, tik-tok needs to move at the edge for that to happen. Nowadays, this seems much more probable than any other time in history, i.e last 15 years.

Personal recommendation algorithms using RL, are certainly possible, because RL is the least computationally intensive training, compared to supervised, unsupervised etc. They also require some kind of social structure, but that will be setup on a blockchain, in an totally open and transparent way, yet private.