Super Data Science: ML & AI Podcast with Jon Krohn avatar

Dexa/Super Data Science: ML & AI Podcast with Jon Krohn

Learn more
Published Jun 23, 2023

690: How to Catch and Fix Harmful Generative A.I. Outputs — with Krishna Gade

Krishna Gade, CEO of Fiddler.AI, dives into the crucial aspects of AI trustworthiness, addressing the challenges of ensuring transparency, explainability, and bias detection in generative AI systems. Discover how Fiddler.AI's innovative solutions enhance AI model reliability, paving the way for safer and more trustworthy deployments.
Episode Highlights
Super Data Science: ML & AI Podcast with Jon Krohn logo

Topics covered

Popular Clips

Episode Highlights

  • Explainability

    Fiddler.AI's explainability algorithms enhance the transparency of AI models by probing them with counterfactual inputs, similar to human questioning. explains that this approach helps determine the sensitivity of each input on the model's output, providing a clear understanding of why a model makes certain predictions 1. This method is crucial for building trust in AI systems, especially in complex environments like large language models. emphasizes the importance of trust, drawing from his experience at Facebook, where understanding AI decisions was vital for both technical and non-technical stakeholders 2.

    The key question that we wanted to answer was, why is it saying that?

    ---

    Fiddler's algorithms aim to make AI more interpretable, ensuring users can confidently rely on AI-driven insights.

       

    Validation

    The Fiddler Auditor tool plays a critical role in pre-production validation by assessing AI models before deployment. describes how the tool probes models with counterfactual questions to identify potential failure scenarios, such as inaccurate or unsafe responses 3. This process results in a robustness scorecard that developers can share with stakeholders to ensure confidence in the model's performance. By simulating human questioning, the Auditor helps organizations decide whether their AI models are ready for production.

    It probes with a lot of counterfactual questions to see what kind of responses it is coming up with.

    ---

    This validation step is essential for maintaining accuracy and safety in AI applications.

       

    Enterprise AI

    Fiddler provides enterprise-specific AI solutions that address the unique challenges faced by large organizations. highlights the importance of AI governance, particularly in regulated industries like finance and insurance, where model risk management is crucial 4. Fiddler's tools help create comprehensive AI validation reports, ensuring models meet regulatory standards and perform reliably. By offering these solutions, Fiddler supports enterprises in adopting AI technologies while maintaining trust and compliance.

    We started Fiddler four and a half years ago with this mission to build trust into AI.

    ---

    These capabilities are vital for organizations looking to integrate AI into their operations confidently.

Related Episodes