Often, computer programming uses Math as a means for describing and then tweaking representations of real-world phenomena. For instance, it is used to represent weather conditions or nodes in an artificial network or represent pixels on a computer screen. Math is used in such instances to perform calculations on matrices. For example, when programming games, matrices help describe all the possible movement options. To realize the movements, the matrices are often multiplied or added or sometimes both actions are performed. All this needs a lot of work, especially when the matrices are larger. This is the reason why computer scientists have spent a lot of time and effort developing highly efficient algorithms to get the job done. In 1969, Volker Strassen, a mathematician figured out a way to multiply two 2x 2 matrices using seven multiplication operations instead of the usual eight operations. Using this new format, researchers at DeepMind are figuring out if it is possible to use a reinforcement-learning-based AI system to develop new algorithms with fewer steps than the existing ones. The team also looked into gaming systems for inspiration, keeping in mind that most systems are based on reinforcement learning. After developing a few systems, the team shifted its focus to tree searching which is also used in game programming. Tree searching enables a system to look at various scenarios based on a particular situation. By using it to multiply matrices, the researchers found that converting an AI system to a game improved searching for the most efficient way to get the desired mathematical result. Additionally, the researchers tested their system by allowing it to search for, review and then use existing algorithms, using rewards to pick the best one. The system figured out the factors that improved multiplication efficiency. Also, the researchers allowed the system to create its own algorithm in the hope to improve efficiency. The researchers figured that in most instances, the algorithms chosen by the system were better than those created by their human predecessors.
