This is a new little series of posts. As a learning technique I’ll be sketchnoting a talk, article or book chapter then sharing the notes and my reflection (even if my sketchnotes aren’t the prettiest).
Yesterday I watched a MoTaCon talk called “GenAI Wars Episode 3: Return of the explorer” by Martin Hynie.
Reflections
It was interesting learning about how exploratory testing has such value when testing AI. The fact that AI defects can be as a result of emerging people behaviours, geopolitical, social or legal in nature was fascinating. The example around an internal build of his financial app recommending selling meth was hilarious yet scary. I’ve seen interesting examples too.
I’d love the opportunity to get to test AI at some point!
Another re-occurring theme was that because of the multiple factors in a failure, doing RCAs was no longer relevant and we should be probing. Interestingly I think this is very similar to my Meaningful RCAs. Perhaps I can rename it meaningful probes? Sounds a bit X-Files to me.
Side note: Amusingly I was writing this whilst listening in to This Week In Quality and ridiculous Gen AI behaviour was a topic!