Have you ever heard the story that your job is automatable, that all the human testers will be replaced by machines or automated tests and you will lose your job? Or even worse, that machines and artificial intelligence will take over our craft and our life and we will be totally useless. Do you buy these? Are you afraid?
“Come with me, if you want to live” – this was the famous line that many members of the Human resistance in the Terminator franchise used, when offering their help in the war against Skynet.
So, come with me (and John Connor), and join the testing resistance to fight on the side of intellect against the evil machine army. I am willing to challenge the I part in AI on contest by focusing on few key topics:
Can we translate testing into machine language? Polymorphic and mimeomorphic actions – what are these?
Do we really know what are the benefits of human testing? What are human testers irreplaceable for?
Do we really have empirical evidence that computers are capable of doing professional testing? Do we have evidence of “intelligence” at all?
Last year at RTC ‘17 I was asked – “Is AI the answer to all test automation problems?”. My answer is “No, it’s not!”. And this talk is my explanation why.
2. Viktor Slavchev
• Software tester
• Blog: mrslavchev.com
• @thetestingtroll
• Insert random brag about my
achievements
• This slide is bullshit
• Nobody reads this far
11. Perfect for the industry –
production line worker
• Simple executable steps
• Easily repeatable
• Easy to measure, control,
understand
• Easy to replace
13. If we were to compete
against machines, what do
we got?
Few key points, that matter about human intelligence in
testing
14. The social nature of
testing
• We are not “just producing code”
• “Quality” and “risk” are social
concepts
- they can not be codified
• The product solves a social
problem
• Just like money
Beating robots down with bare hands -
Quality
16. Action vs. behavior
• Behavior – roughly equivalent to
“internally produced movement or
change”
• Action – human actions are species of
behavior, but the key component is they
have intent.
Harry Collins – “The shape of actions”
17. Example: wink vs. blink
• Blink – involuntary action of closing the
eyes.
• Wink – Voluntary action with intent to
send a message, depending on social
context.
18. Mimeomorphic and
polimorphic actions
• Morph – shape
• Mimesis – imitation
• Mimeomorphic - actions that we want
to do the same way every time
• Polimorphic actions - are naturally and
appropriately variable, and are rooted
in social and human interactions and
goals
Michael Bolton – Shapes of actions
19. Example in testing
• Mimeomorphic action – following
a test script, procedure, scenario
• Polymorphic action – designing a
test case, exploring product
requirements, designs
• Human actions are naturally
variable
• Machines are efficient for
producing mimeomorphic actions
28. When somebody says your job is
automatable/AI-able
Your expertise Their idea of your expertise
29. Pessimistic prediction
Automation and machines can not replace human expertise, they can
only replace “machine like actions” that we intentionally perform, BUT
• There’s a negative trend of deskilling human expertise
• Making it look simplistic, robot-like, procedural
• In testing – to test = script, execute simple steps
If we don’t change this trend, we allow deskilling of human expertise,
therefore automation and replacement.
30. Realistic prediction
We can use automation and AI to
support testing by using them as
counseling expert systems
• To do exhaustive brain work
• To explore exhaustive scenarios
• To aggregate and evaluate
quantitative data
31. Trollistic prediction
• None of this is true
• Everything is a lie
• All the automation/AI hype is a
decoy
• To draw our attention from the
real problem
• That our jobs are going to be
taken by…
• DOGGIES!!!