In testing, can AI and machine learning deliver with accuracy?
The appeal of artificial intelligence is understandable, especially as it applies to the more tedious or laborious of IT tasks. Software testing surely seems like the type of function that would benefit from AI and machine learning.
With AI in software development -- and specifically in testing -- what's uncertain is how accurate those tools can become relative to what human testers can accomplish. It will take time, effort and lots of data to create AI that's smart enough to reliably test software, argues software consultant Matt Heusser in this handbook's featured article. An important element in that process, he explains, is how humans will teach automated tools to know what it means for code to be correct. This is a formidable challenge, and it's difficult to imagine how the hurdle can be cleared in all -- or even many -- cases.
Testers will need to find the situations where machine learning and AI can prove effective. The use of AI in software development isn't about pushing aside humans. Instead, it's about how developers and testers identify the right scenarios for AI to be accurate and efficient enough to earn its place alongside the people tasked with this important work.