Reliable cognitive solutions
Traditional software is built using hiding and the decomposition process. The system is broken into sub-systems which are then broken into components and so on until the lowest object in the hierarchy is defined. With the introduction of components implemented using machine learning and dependent on data a host of new challenges to the system readability are introduced. Challenges include a good match of business objectives, machine learning optimization objectives and relevant data, stability of the quality of the solution due to changes in data, artistic choices of machine learning parameters as well as correctness of components in probability. In this talk I present an end-to-end view of the challenges and possible ways to address them as well as some resulting interesting research directions.