Artificial intelligence and discussion surrounding it are neither new nor surprising. However, the fever pitch regarding AI has never been stronger as the current times. It appears, for the first time, that viable AI solutions for what was previously science fiction will become available.

How do we test AI applications, is still a big question.

This talk explains the basic AI program development lifecycle and possible involvement of testers in testing the AI applications. Explainability – Why did the AI take the decision it took?  This forms an important part of AI testing.

This talk will cover that question, as well as metamorphic testing and methods developed by Vipul, for creating tests for AI which can also reveal bias in the data. 

Key Takeaways

• Key challenges in testing AI applications
• Explainability of AI results and tools to use for it
• Technique to create tests for AI
• Metamorphic testing