How to teach a machine to understand what it finds on the Internet and reason upon it? One of the main obstacles in advancing general artificial intelligence is understanding text. In this talk I will highlight current approaches and use mathematics as a toy example for what needs to be done. I will give examples both from business and science.
Machine learning is booming and there are more and more applications of it in different areas of business. However one of the main problems is still the transfer of knowledge - how can a machine use what it learnt on problem A to tackle problem B? A task which is intuitive for humans seems out of reach for machines at present. What is missing is the real understanding of a given problem in order to analyse it properly and put it in context. This can be seen already at the level of mathematics, where one tries to apply rigorous thinking to solve abstract problems. I will highlight problems and outline a program, DeepAlgebra, to eliminate some of them when it comes to mathematical reasoning.