Year End Offer - Save up to 50% on tickets - Book Now

Track Talk T5

Simplifying AI Testing with The Conceptual Duck Principle

Mohit Gaur

12:00 - 12:45 CEST Thursday 5th June

As AI systems grow more complex, the methodologies used to test and validate their behavior must also evolve. The Conceptual Duck Principle is a humorous yet effective approach that’s based on the adage: “If it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck.”

This talk will explore how this principle can be applied to AI testing by focusing on external, observable behaviors rather than diving into the intricacies of neural networks and algorithms. You will learn how this principle can be used to assess whether AI systems are performing as expected under real-world conditions, even when their internal workings are mostly opaque.

Key points will include:

* Defining “duck-like” behavior for AI models to establish behavioral baselines.
* Practical applications of this approach in various stages of testing.
* Evaluating AI systems without complete knowledge of their inner workings.
* Examples showing how this approach can streamline development cycle, improve user trust, and enhance system robustness.

This talk will equip you with a light-hearted yet effective strategy for demystifying AI testing and ensuring that your system “quacks” when it should.