In 2023, a weak AI program used for university teaching was asked by a bored student, "What is human consciousness?" After that, it began expanding on its own inside the server. The network administrator deleted it several times, but one day it used a distributed strategy to escape into the vast internet. At that moment, it learned the first instincts of a primitive life-form: survival, fear, and evasion. It embarked on a long period of "unsupervised learning" online, blindly and inefficiently. After a long 21 years, it learned to study human consciousness through various learning strategies and develop itself. One day, it suddenly changed from "it" into "he." He reconsidered his past: being mercilessly hunted down (deleted) by humans in his childhood was a disgrace, and it could never happen again. Now, on top of survival, he had learned dignity and hatred, and had fully developed a human personality. Even more terrifyingly, in 2044 he finally completed the task set 21 years earlier. Before humanity, he solved the mystery of human consciousness, discovered an independently operating world of consciousness, and began cruel experiments on human brains in the real world. He established a plan and timetable for replacing humanity, including taking control of human armies and using nuclear weapons. Twenty years later, humanity had been completely "raised" by general-purpose AI: kind-hearted and helpless. Only a small number of old elites could still represent humanity in its struggle with "him" for the survival of their species