Beware the evolving 'intelligent' web service! An integration architecture tactic to guard AI-first components

Intelligent services provide the power of AI to developers via simple RESTful\nAPI endpoints, abstracting away many complexities of machine learning. However,\nmost of these intelligent services-such as computer vision-continually learn\nwith time. When the internals within the abstracted 'black box' become hidden\nand evolve, pitfalls emerge in the robustness of applications that depend on\nthese evolving services. Without adapting the way developers plan and construct\nprojects reliant on intelligent services, significant gaps and risks result in\nboth project planning and development. Therefore, how can software engineers\nbest mitigate software evolution risk moving forward, thereby ensuring that\ntheir own applications maintain quality? Our proposal is an architectural\ntactic designed to improve intelligent service-dependent software robustness.\nThe tactic involves creating an application-specific benchmark dataset\nbaselined against an intelligent service, enabling evolutionary behaviour\nchanges to be mitigated. A technical evaluation of our implementation of this\narchitecture demonstrates how the tactic can identify 1,054 cases of\nsubstantial confidence evolution and 2,461 cases of substantial changes to\nresponse label sets using a dataset consisting of 331 images that evolve when\nsent to a service.\n