Software Quality for AI: Where we are now?
Authors: Valentina Lenarduzzi, Francesco Lomio, Sergio Moreschini, Davide Taibi, Damian Andrew Tamburri
Artificial Intelligence is getting more and more popular, being adopted in a large number of applications and technology we use on a daily basis. However, a large number of Artificial Intelligence applications are produced by developers without proper training on software quality practices or processes, and in general, lack indepth knowledge regarding software engineering processes. The main reason is due to the fact that the machine-learning engineer profession has been born very recently, and currently there is a very limited number of training or guidelines on issues (such as code quality or testing) for machine learning and applications using machine learning code. In this work, we aim at highlighting the main software quality issues of Artificial Intelligence systems, with a central focus on machine learning code, based on the experience of our four research groups. Moreover, we aim at defining a shared research road map that we would like to discuss and to follow in collaboration with the workshop participants. As a result, the software quality of AI-enabled systems is often poorly tested and of very low quality.
Unternehmen: LUT University