Technical Risk Analysis for AI Systems - Bill Matthews

Hardly a week goes by without a report of some demonstration of the Unintelligence of AI systems; some are funny and harmless but others more serious and harmful. Most testers are already aware that AI systems can be challenging to test and the evidence of Unintelligence in such systems suggests that we don't yet have a clear model of how to approach testing of AI system.

In this talk I will share a framework for examining Risks (Business, Ethical and Technical) related to AI Systems that I have been developing as ways for testers to think critically about what is important to test. In particular, this talk will focus on the Technical Risks that come from the AI Architecture, Data and Learning.

Takeaways:

• Understand the breadth of the Risks from the Business, Ethical and Technical viewpoints
• Understand in more detail the Technical Risks
• Strategies and tactics for testing Technical Risks in AI systems