We have witnessed AI agents that can beat World Chess Champions and Go, Masters, recognise faces, place and objects, and even write articles. However, despite the best efforts from the research community, no AI model has been trained successfully to do symbolic reasoning tasks, which are involved in advanced mathematical operations such as integration. However, the latest research by Guillaume Lample and François Charton, at Facebook AI Research in Paris, claims to have developed an algorithm that can overcome the barrier. 

Lample and Charton, in their paper titled "Deep Learning for Symbolic Mathematics", explains that AI agents "can be surprisingly good at more elaborated tasks in mathematics, such as symbolic integration and solving differential equations," contradictory to the existing beliefs. The pair "trained a neural network to perform the necessary symbolic reasoning to differentiate and integrate mathematical expressions." Which is a "significant step toward more powerful mathematical reasoning and a new way of applying neural networks beyond traditional pattern-recognition task," reports MIT Technology Review. 

Their success lies in unpacking mathematical shorthand into its fundamental units, which is a critical element in advanced mathematical operations. Then taught the neural network to recognise the patterns of mathematical manipulation that are equivalent to integration and differentiation.

The researchers claim that their model results that outperform existing commercial Computer Algebra Systems such as Matlab, Maple, or Mathematica.