Bring the tester’s toolbox to the world of AI-based systems

Track B - Outer Space theme

Systems and software based on artificial intelligence is becoming a more and more natural part of our society and our daily lives. Still there seem to be a hurdle for traditional software testers to cross over and to start exploring this new world. In this paper I would like to lower that hurdle by looking into what the main challenges are in testing an AI based system compared to a traditional system. I would also like to dig into the tester´s existing toolbox to see what tools (skills, methods, techniques etc.) that we have that we can bring with us into data science and machine learning and that can use to add value. With the black box approach, we do not need a too detailed knowledge of the structure and components of the AI as long as we are familiar with its characteristics and behaviour. As a final round I will investigate the tools (skills, methods, techniques etc.) that needs to be upgraded or added to give the tester full confidence in transitioning between traditionally developed software and systems into systems constructed by one or several layers of AI.

 

Key takeaways

  • Challenges testing AI based systems
  • How testers bring value in development of AI based systems
  • Tester’s toolbox already full of useful tools