Applying AI to testing
summary
Claims that combining a trainable monkey tester with AI and a difference testing oracle lets random exploration generate maintainable functional regression suites, not just crash reports, and cites testing as ~30% of a budget. Uses Selenium, pixel/approval comparisons, gremlins.js/Chaos Monkey and Android monkey examples, and implements a Java Swing prototype with a whitelisting approach to ignore expected changes. The prototype runs and produces executable test cases; humans still decide regressions.
AI-written summary · about
chapters (14)
- 0:00Intro
- 1:17How testing works
- 2:03The Oracle Problem
- 2:44AI in Testing
- 3:34The Goal of Testing
- 6:47No bugs only surprises
- 9:03Regression testing is change control
- 16:30Monkey Testing
- 17:07Monkey in Action
- 19:09Smart Monkey
- 20:11Dump Monkey
- 22:11Jobs to go first
- 24:14Should you be afraid
- 26:12Summary
related talks
01h 01mRise of the Machines: Can Artificial Intelligence Terminate Manual Testing?
41mHow AI is transforming software testing
44mHow AI is Exposing the Existential Crisis of Software Testing
32mQuality engineering in the AI age: From testing AI to leveraging GenAI
01h 02mReduce Developer Toil: Autonomous AI for Enterprise Testing @ Scale
40mFrom Bugs to Brilliance: Leveraging AI in Modern Software Testing
50mPanel Discussion: What Can Go Wrong With AI in Testing?
32mWhy testing is a developer's problem