Hi MoT community!
I thought I could start a new thread about challenges and appliances of machine learning in testing. We are building autonomous testing tool called AutoExplore. The tool can currently browse any Web UI fill forms etc. However I would like to take it’s testing capabilities to next level.
The tool uses LLM for generating test data to submit through the UI, this seems to work okay. The data seems realistic enough for test usage.
I have tried scanning images, using clever prompts and prompt context’s telling the LLM to test like professional test engineer and analyze the image for any errors. However this approach seems to cause many false positive results or responses which often are matter of opinion.
- Counting of things seems to often go wrong, even for plain simple UIs and clear images
- It can recognize text contrast issues quite well, but that can be tested more efficiently using ready made tools
How are you using AI / LLM in testing and where do you see most benefits?