How does this Todai Robot passes a multiple choice entrance exam with more that 85% accuracy and then writes an essay on a given topic? And what makes human brains stand out when compared to AI?
This post is a short summary of a TED talk given by Noriko Arai, AI expert from Japan. She is the programme director of an AI challenge that evolves around the questions of whether AI get into the University of Tokyo and to determine the limitation and opportunities of AI by setting a specific goals to be achieved – passing the university entrance exam in this case.
In Japan, students are usually required to take a two stage entrance exam to get admitted to university. First rounds is a multiple choice exam. Upon passing that exam, students are asked to write an essay on a given topic. I bet you can relate how the average student prepares for these test, but this is how AI Robot Todai does it.
Todai is not able to read, but he is amazing when it comes to searching and optimising. Imagine AI is asked “Mozart’s last symphony shares its name with this planet”. AI compares keywords from the questions with content on the internet and recognises that keywords like “Mozart,” “last” and “symphony” are heavily co-occuring with a name of a planet. Based on that he finds the answer (which is Jupiter btw). Furthermore, Mrs. Arai explains how AI writes essay better than some students and solves mathematical problems written in natural language. However, besides all these AI success, there is one “big flaw” when it comes to AI: AI does not understand meaning. Despite accumulating more that 15 billion English sentences through deep learning, they make errors when asked questions that require understanding of a situation. In such situations, AI disguise as if they understand, when in reality they don’t.
Arai finishes her talk with an appeal to initiate a revolution in education. Arai (2017) describes how “most of the students just pack the knowledge without understanding the meaning of the knowledge, so that is not knowledge, that is just memorising, and AI can do the same thing”.
This final thought of hers goes hand in hand with the article The Human Brain vs. Computers in which van Paasschen (2016) elaborates why the human brain still has a leg up on AI. The human brain possesses powers such as parallel processing, sorting information and tell jokes or respond in a sarcastic manner (Van Passchen, 2016), nevertheless pace of ever-improving technology does not seem to slow down and smart machines make themselves smarter and smarter.
In essence, both the talk and the article explain what makes our human brains outstanding compared to AI, but who knows if AI will tell us sarcastic jokes and give meaning to its knowledge in the future. What do you think?
Arai, N. (2017). Can a robot pass a university entrance exam?.
Available at: https://www.ted.com/talks/noriko_arai_can_a_robot_pass_a_university_entrance_exam#t-188250 [Accessed 18 Oct. 2017].
Van Passachen, F. (2016). The Human Brain vs. Computers – Thrive Global. [online] Thrive Global. Available at: https://journal.thriveglobal.com/the-human-brain-vs-computers-5880cb156541 [Accessed 18 Oct. 2017].
