With the latest openAI trendy about āChatGPTā, Iām curious to know your opinions about it, do you think it could improve your testing tasks?
In my opinions, it could be effective in testdata generation, in devops pipeline technical tasks, maybe giving new testing ideas that could be added to exploratory testingā¦
I also have more questions for you,
How can we become more productive using its hacks and (sometime smart) replays ?
What are the new tasks that could be easily automated with chatGPT ?
What new challenges we could encounter in the future as testers ?
Shall we also include them into our testing strategy?
Iāve used it to read the long article for me and provide me a short summary, itās fairly accurate but a little too literal sometimes, but still it seems useful. Maybe the next thing to try is to have it generate some test data!
Looks like the sort of progression I was using to create test datasets 25 years ago when I was working in utilities regulation - though those were for numbers like āLength of water mains refurbishedā and āNumber of properties connected to mainsā, and the progression was over time. And I knew what the likely upper and lower limits of each data point were.
This got interesting where one client started falsifying their data returns and were providing numbers with similar progressions to my test dataset as their live data. As real-world pricing decisions were based on this data return, there were consequences (a Ā£35 million fine and an investigation by the Serious Fraud Office).
Which suggests a use for an AI system, albeit more in validation than in testing: using ChatGPT output as a benchmark against which future data submissions might be compared. If the progression patterns within the submitted data was too close a match to a ChatGPT ābenchmarkā dataset, the system would raise a flag. Especially useful if the dataset has something like 40,000 data points, like the one I was working on.
thatās a good point ! talking about blogs Iām also thinking to make a serious interview with chatGPT and then comment about the result in a funny way
BTW, chatGPT has funny jokes to tell
I used it to help me get started with the code for API tests, it worked pretty well actually. Not sure how it would work for complex scenarios, though.
In terms of code, itās been working pretty great for things thats more commonly known. I used it to help me write k6 tests for an AWS redshift cluster.
It gets a bit complicated when there isnāt a lot of resources about a given topic. My recent experience left me with library functions that doesnāt exist
A reminder : do not use chatgpt to analyse customer sensitive data in any way, or any feature work that is not public. it will leak that work for you. Even grammar checkers will leak your little snippets of documentation you wrote and just wanted help correcting typos in, back into the web via itās learning set. Any code it writes may well see you getting into trouble for stealing code (assuming the code works), because of the same learning set leaking behavior.
That said, itās a great tool for bouncing ideas of of. And great for researching , often better than google.
Itās actually good to give several examples to explain certain things in the testing world.
Itās good for remembering frameworks, what I mean by this is letās say I havenāt used JMeter in a year or 2 and I need to start using it again.
And you ask the question to ChatPGT āHow do I make a performance testing script in JMeter?ā
It comes up with a very good direction and explanation of how to do it.
You can paste an API request in it and ask how to test it. You can keep asking āis there anything else you could testā and it will provide you more
! 45 day or month of birth 
