I am going to challenge some of the ideas in education part of the report. 

Based on the discussion in education section, one of the main reason for the current absence of sophisticated use of AI technologies in schools is lack of data establishing the technologies effectiveness. And some online learning systems like MOOCs can be considered as vehicles for collecting such data, and testing the new education technologies. Though the report mentions that the capabilities of MOOCs have been overestimated, I think another possible problem is that the data collected in MOOCs might be biased. It means the data collected this way cannot capture the feature of the whole population and it would possibly lead us to a wrong direction.

There are mainly two types of students, one is the students who are interested in the course and willing to learn and the other is those who are not involved much in the course. For those who have less engagement in the course, or who care more about earning credit rather than learning, a better educational technology is supposed to be inspiring, easier for understanding and better at attracting attention. And for the students who are interested in the course, they would benefit more from deeper or wider topics. However, people taking online courses are more likely to be people who are willing to learn. 

In addition, it is more likely that the users of MOOCs and other online courses are adults rather than school, college or university students. Adults and children might use different ways to learn new stuff. For example, adults might need logical explanation for better understanding but children would prefer more examples to understand the situation. And a colorful design for material with more pictures might attract children’s attention and help them involved in the course, but it might be more efficient for adult to contain jokes and provide clean-layout material. So they adult and children would probably benefit from different tutoring systems.

Thus, the educational technologies that are analyzed to be well-performing in MOOCs might not be suitable for general classes.

Besides the biased data problem, there is another possible opportunities for AI application in education. A traditional class is usually conducted by a teacher or professor and the whole class would share the same pace and materials. Based on the report, if we can analysis how good a student is performing with the help of AI or machine learning, and provide him materials accordingly, it is possible that the student can get help and mentored individually. An AI evaluation system can not only help with students individual learning, it can also be used as a reference for final grades. I believe an AI system is more reliable comparing to homework and “one-time” final examination assessment. They can not only evaluate how well a student performs in the class, but also detect how much effort the student put into the course, for example how often the student respond with the material or exercises provided by the AI system.

