The article Artificial intelligence and life in 2030 reviews the current state of artificial intelligence research and explores its application in eight different fields. One of which is education. It argues that artificial intelligence is able to improve all levels of education, from K-12 to university students, by offering a personalized content (pg. 31). While this sounds promising, applying artificial intelligence in education can be a lot more challenging because insufficient data and restriction on the variety of question format make application of artificial intelligence in education less effective and hard to popularize.

First of all, lack of data can greatly decrease the effectiveness of application of artificial intelligence in education and thus might lead many students to choose traditional learning methods over it. Machine learning, which is what the article argues to provide personalized curriculum to students, relies on large data set (pg. 31). In order to provide an individual student with personalized content, machine learning requires data from this individual to train the model. In other words, instead of training artificial intelligence with data from every students using this system, only data collected from a particular individual can be used for training. The requirement of large dataset is a limitation of machine learning. In some circumstance, this limitation is a less serious problem. For instance, machine learning is used in a recommendation engine to recommend users of an online shopping website products he or she might be interested in. In this scenario, a somewhat inaccurate result set is acceptable. However, when it comes to education, this can be more problematic. With limited data, the artificial intelligence is very likely to have the student repeat something he or she already knows very well while not offering enough practice for what actually needs more attention. A direct result of this is an ineffective education result and an unpleasant learning experience for students (for many, being forced to repeat things one knows very well of is not interesting). If students feel the artificial intelligence teaching system is not as effective as they expect, they could quickly lose patience and switch to more traditional learning methods such as flash cards. The article argues that the reason artificial intelligence is still not widely used in colleges and schools is a lack of funds and records supporting the effectiveness of the technology (pg. 33). But this is not the core issue. Unless the limitation of data set is taken care of, it will be hard to attract funds when students do not want to use it and there will be no such records because it is actually not effective.

Another challenge of applying artificial intelligence in education is that it restricts the variety of question formats since it cannot effectively read answers students input. The article mentions that progress has been made about natural language processing to understand widely spoken human speech (pg. 9). Right now, with natural language processing, artificial intelligence like Siri is able to do some speech recognition after being trained with a large amount of data. However, written answers is different from spoken language. It may contain figurative language (such as paradox, metaphor) or technical/scientific information. Artificial intelligence is not capable of knowing whether these sentences are correct (close enough to the real answer). Therefore, artificial intelligence teaching system can only have questions in a format like filling in the blank or multiple choice, which is not always the optimal format in terms of effectiveness.

Reference (short for the reason of essay length):
Artificial Intelligence and Life in 2030 C One Hundred Year Study on Artificial Intelligence, Sep. 2016
