Current AI hype needs a reality check. Most researchers are still stuck in 2019, people are getting too caught up in transformer models, and no one's really exploring the real potential of graph neural networks.