AI can’t mark its own homework

Artificial intelligence is rapidly changing how software is designed, written and tested. Development teams can now use AI to generate code, produce test cases, identify likely defects and automate repetitive quality assurance (QA) tasks at a speed that would have seemed unrealistic only a few years ago.

That acceleration is valuable. But it also creates a new QA problem.

CEO of T-Plan.

When the same class of technology is used both to create software and to decide whether that software is correct, organizations risk building a closed loop of confidence. An AI model may generate code based on a particular interpretation of a requirement, then generate tests based on the same interpretation. If the original assumption is wrong, both the code and the test can agree with each other while still failing the user.

AI, in other words, cannot be the sole judge of its own work.

This is not...

Copyright of this story solely belongs to www.techradar.com. To see the full text click HERE