“The development of large language models like GPT-4 has inherently solved some alignment problems, such as intent alignment, by requiring them to develop a common-sense understanding of natural language.”