736: How to Officially Certify your AI Model — with Jan Zawadzki

AI Safety

Jan Zawadzki explains the importance of AI safety standards in the automotive industry. He highlights how standards like ISO 26262 ensure that AI systems in self-driving cars are reliable and safe, preventing potential fatal accidents 1. Jan's experience at Volkswagen's software development subsidiary, Cariad, involved creating processes and tools to implement automated driving functions safely.

In the automotive industry, you don't need a government agency to tell you, hey, you have to be sure that your AI system does what it's supposed to do. We know that we do not want like anything that can go wrong in autonomously driving cars. It can kill people, literally.

--- Jan Zawadzki

He now aims to apply these safety measures to other industries, such as healthcare and manufacturing, through his work at Certif.ai 1.

   

Risk Fairness

Jan discusses the methodologies used to assess and mitigate risks in AI systems, ensuring fairness and reducing biases 2. He emphasizes the importance of evaluating risks like discrimination based on ethnicity, gender, or other factors, and implementing measures to mitigate these risks. This approach is crucial for creating AI systems that are both fair and reliable.

No system will ever be without risk. But by this approach, by defining measures to mitigate those predefined risks, we can create an acceptable risk that we have at the end.

--- Jan Zawadzki

Jan's methodology includes examining the data used for training, the models employed, and the deployment process to ensure that AI systems remain robust and fair over time 2.

   

AI Certification

Jan highlights the role of AI certification in improving the quality and reliability of AI systems. He explains that certification processes help identify and rectify suboptimal engineering practices, ultimately leading to better AI products 3. This is particularly important in high-risk applications, where trust in AI systems is crucial.

What motivates me, I'm motivated by improving the engineering quality of AI based systems.

--- Jan Zawadzki

Jan's work at Certif.ai involves testing AI applications and preparing for upcoming regulations like the EU AI Act, which will require external auditing for high-risk AI applications 4.