Reinforcement learning, when applied with human feedback, performs worse on probability and falls more in line with human reasoning. While the addition of transformers is not guaranteed to result in artificial general intelligence (AGI).