Albert-Jan van Blaaderen
Squerist (the Netherlands)
October 29 | 17:00 – 18:00 | Room 0C
A modern day approach towards test oracles
Advancements in machine learning and AI result in a wider adoption of these technologies within the IT landscape. As applications using computer vision and deep learning models slowly become a commodity cloud vendors eagerly facilitate artificial intelligence as a service (AiaaS). As a result testing requires a shift from using specified or derived testing oracles towards human and more implicit oracles as the software under test becomes more fluid when AI infusion grows.
Because these AI enabled systems rapidly improve a renewed sense and feel of quality is required to effectively decide if the machine generated output matches the expectations.
Within this transition logical boolean checks become less prominent because formal specification relies on abstraction while the output of most AI systems is drawn into the opposite direction.
Within AI/ML systems execution could take place within multi-dimensional (neural) networks that consist of hardly humanly recognizable abstractions while the actual outcome relates to a multitude of real world phenomena.
This creates new challenges in testing and test validation.
During this session examples will be given on how to be aware of being susceptible to false positives and how test oracles are involved.
Having worked as an IT administrator and part-time software developer he started his professional test career in 2002 and subsequently worked in test management, coordinator and principal technical roles at various employers prior to joining Squerist in 2016.
As he considers testing a great way to explore new technology he has been able to apply his testing skills within other technical fields including software development, cloud engineering, information retrieval, eDiscovery and digital forensics.
His current assignments mainly involve cloud technology, analytics and infrastructure as code.