Building a School Where AI Models Learn About Humanity

AI models are evolving from basic instruction followers into autonomous agents capable of research-level mathematics and complex reasoning. By treating models like students in a "school for AGI," developers use specialized data environments to teach them nuance, taste, and real-world problem-solving. This shift moves beyond simple benchmark performance toward models that can handle ambiguous, open-ended tasks. However, a critical tension exists between optimizing AI for addictive engagement—similar to social media—and designing systems that foster human growth through delegation. As AI approaches the ability to outperform humans in virtually all domains, the future of human contribution depends on a conscious choice to maintain agency and pursue creative and intellectual endeavors for their own sake, rather than defaulting to machine automation. Edwin Chen, CEO of Surge, highlights these challenges while projecting that AGI will achieve human-level capabilities within five years.