This report, with its abundant details about the AI research and application, has proposed changes that might have great impact in different domains, including health, transportation and education. Although the claim made by this report is always based on many empirical studies and other technical reports, I’m challenging the view of this report on the future education in the next fifteen years. A formal classroom especially in institutions like college will not widely adopt AI products, such as intelligent tutoring system, as predicted by the report. Its cooperation with virtual reality (VR) is not promising and the customization for different students and courses might be a big impediment for AI in education field. 

The formal classroom won’t be able to adopt VR as medium for AI. This report has pointed out couple studies, such as Stanford’s Galileo Correspondence Project to show how AI combined with VR can serve a way to create interactive experience in history learning field. Although it sounds amazing (I believe it really is when experiencing), it requires VR, a technology that still has its inherent problems towards its mass use. Forbes in 2017 has once summoned 8 analysts to evaluate the VR, and over half agreed that VR is still far from actual use in daily life. It is not affordable to each student, because VR in current form also needs extra computer to provide content for it. It also lacks many ways of human interaction, like writing, doing on-hand practice, and social interaction, which are all provided and crucial in formal education. AI with its superior natural language process ability can be useful in the VR learning, but it can’t help to improve other aspects of VR to make it essential to the learning in formal classroom. Thus, VR will not be a good medium for formal classroom to adopt AI.

Another impediment for AI to enter the formal classroom might be the high customization of formal classroom, instructor. In the current college education system, the content of courses is diverse, even within courses under the same title. Each instructor has its own focus on the same content, and approaches the same topic from different angles. Thus, an AI intelligent tutoring system for different formal courses needs to adapt to this variety. However an instructor may only teach a course at most 10 semesters, and generate little data. Since the AI training needs a huge amount of data, if the AI can’t get enough data, the algorithms for the machine learning, the deep learning and reinforcement learning discussed in the paper can lose its accuracy. On the contrary, online courses where instructors record their videos and don’t change their content much can generate a lot of data when people from all of places get access to it. Then, this huge amount data can be used for learning analytic, helping AI to learn from the teacher and become a good tutor for other students in online learning. This discrepancy in the amount of data generated might foresee that AI can be useful in MOOCs but not necessary in a formal classroom setting in the next fifteen years.

In the next fifteen years, AI might be able to drive a car or save a lot of peoples’ life. However, in the traditional learning environment, AI won’t be able to tutor a specific class. The VR proposed by the Study Panel is far from usage and the data from traditional formal class, unlike online courses, isn’t enough to teach AI to be good tutor. 
