Title: How to Make Machines Understand Math?
Abstract: Since the breakthrough of Google's AlphaGo
program that defeated the world Go champion in March 2016, the
Artificial Intelligence (AI) technologies, mainly machine learning,
become very popular in the technology world. Most applications using
the machine learning technology heavily rely on certain statistics
models, which may not be very effective in solving math problems. In
this talk, a novel method called experience-based approach (EBA) is
developed to solve math problems, which replaces the traditional
axiom-based approach (ABA). The main advantage of the EBA method is
that the solutions generated are human-readable, while the
ABA-generated solutions are only readable by experts with special
training. The human-readable requirement is essential in education.
Another important consideration in education is that the system should
help students understand math, not simply providing them solutions
without any math understanding. In order to develop such a system, we
first need to make machines understand math, which means that we need
to create a mechanism to represent math expressions so that the pattern
matching technique can be applied in solving math problem. A demo will
be given to show how the system can solve a simple math problem with a
sequence of suggestions. The students using the system still need to
make their own decisions based on their math understanding.