1) People want testers AND machines to "test a product as it's actually used". Nah. If that were so, not only would you NOT provide them with explicit instructions like "do this, then this, then this"; you'd tell them to act like a real user and ignore any and all documentation.
2) Claims that automated checks at the GUI level "simulate what a real user would do" are err in at least two ways: first, people don't act like machines; second, people don't have the capabilities that machines would afford if only we stopped trying to make them act like people.
3) My experience and epiphanies this week with co-teaching Rapid Software Testing Applied include this: in a world of ever-increasing application of AI and ML (however bogus it may be), rote automated output checking is an even weaker idea than it's always been.
4) Are you a tester? Do you want to get into "test automation"? Then focus on two things: i) learn to program generally; don't worry about Selenium or Cypress or GUI frameworks. ii) Learn to observe, analyze, think critically, perform experiments. Then use code to enable THAT.
5) That's because, in the face of AI and ML and *all that stuff that has outputs that we didn't anticipate*, our task will not be "given this input, is this output correct?" It will be "given this input, why the hell did it come up with THAT output? Is that what we WANT?"
6) The real upcoming challenge for testers is the same as it's really always been: is the product we've got the product we want? The software business has been fooling itself for almost 60 years by asking "given this input can the machine produce the same result every time?" Duh.
