Intelligent Tutoring Systems (ITS) simulate the behavior and pedagogy of human tutors. Several meta-analyses have found that ITS are generally as effective as human tutors at promoting learning. Unfortunately, ITS are extremely expensive to produce, with some groups estimating that it takes 100 hours of authoring time from AI experts, pedagogical experts, and domain experts to produce 1 hour of instruction. The expense of creating ITS seems to be the largest barrier to scaling up ITS for widespread adoption. Some groups have created specialized authoring tools to address this problem. However, these authoring tools still require experts to use them. We have developed an alternative approach that replaces authoring tools with a new learning environment and replaces experts with novices. In our approach, novices read static content like books and web pages together with a virtual student who proposes summaries, questions, concept maps, and predictions about the content being read. The virtual student combines AI with the corrections of previous human novices to continuously improve its summaries, questions, concept maps, and predictions. Our previous research has shown that by correcting errors that the virtual student makes, human novices learn the content better than reading alone. Moreover, by correcting errors, human novices implicitly author the content needed to create an ITS. Our current research has created the infrastructure to implement this approach in the real world and has evaluated this infrastructure by generating over three thousand tutoring modules from the Navy Electricity and Electronics Training Series. The tutoring modules run on a previously developed ITS for electronics that tutors students by holding a conversation in natural language.