The author initially believed that GPT-4, a state-of-the-art AI text generation model, had degraded in performance, but has since revised this opinion, acknowledging that their initial tests were flawed. They also express frustration with the discourse surrounding AI, criticizing those who make exaggerated claims about its capabilities or fear its potential power.