Summary

Microsoft's on-demand webinar discusses the practical application of Responsible AI principles, emphasizing reliability and safety, and introduces open-source tools for error analysis and model updates.

Abstract

The article highlights Microsoft's efforts to operationalize Responsible AI principles through an on-demand webinar featuring insights from Dr. Besmira Nushi and Dr. Ece Kamar. It emphasizes the importance of reliability and safety in AI systems, the need for detailed model performance evaluations beyond aggregate metrics, and the development of open-source tools to support these goals. The webinar also covers the use of Error Analysis for diagnosing AI model failures and introduces BackwardCompatibilityML for informed decision-making during model updates. The overarching theme is the commitment to ensuring AI systems are fair, transparent, empathetic, and robust, aligning with the four pillars of Responsible AI: fairness, transparency, empathy, and robustness.

Opinions

The author believes that Responsible AI is crucial for organizations to ensure AI decisions are made safely, reliably, and in an explainable manner.
There is a strong emphasis on the responsibility of AI developers and users to eliminate bias and ensure AI technology decisions are transparent and accountable.
The author suggests that the fight for social justice has brought attention to the potential for AI to cause prejudice and discrimination, necessitating a proactive approach to mitigate these issues.
The article conveys the author's view that Microsoft's webinar provides valuable insights into the practical implementation of Responsible AI principles in a large industrial setting.
The author endorses the webinar as a resource for AI engineers and developers, considering it a glimpse into a long-term vision for integrated responsible AI tools.
The author recommends the webinar for those committed to customer-centric and human-centric AI development, indicating a belief in its relevance and utility for the AI community.
The author promotes their own ebooks on AI, suggesting a personal investment in educating others about AI principles and practices.
The author clarifies that there is no professional relationship or reward from Microsoft for promoting the webinar, affirming an unbiased opinion.

How to bring the Principles of Responsible AI to the practice, according to Microsoft.

Microsoft launched an on-demand webinar on how the company brings the AI principle of reliability and safety to reality.

Some weeks ago, I published an article called “A Quick Introduction to Responsible A.I. or rAI,” where I explained the basics of this domain and how rAI can ensure that these decisions are taken safely and reliably and proven explained fashion.

As explained in that article, Responsible A.I. is an essential standard that organizations must achieve because, almost universally, A.I. models can have a tremendous impact on someone’s quality of life.

The fight for social justice in the media is a spotlight on possible causes of prejudice and discrimination situations that A.I. can cause, and it means that whoever develops and uses A.I. should do everything possible to eliminate prejudice (bias), explain how their technology makes decisions, and accept responsibility if A.I. becomes dishonest.

Following a not surprising huge interest from people in my article about rAI, I decide to publish an article called “Introduction to the 4 Principles of the Responsible A.I. for Business Leaders.” where I explored a framework to ensure that organizations understand and apply models that adhere to corporate governance principles like fairness, transparency, Empathy, and, finally, robustness.

These concepts were presented as the four pillars of Responsible A.I.

Going deeper… on Responsible A.I.

To add a further deep-dive on this topic, last week I’ve completed an exciting on-demand webinar promoted by Microsoft about how Microsoft and other industry leaders have made it a priority to deploy A.I. reliably and responsibly.

It is related to the fact that Microsoft has recently developed A.I. principles and standards and a company-wide ecosystem to guide responsible A.I. implementation.

In this webinar, two researchers: Dr. Besmira Nushi, Principal Researcher in The Adaptive Systems and Interaction Group at Microsoft Research, and Dr. Ece Kamar, Senior Principal Research Area Manager at Microsoft Research Redmond, share some critical insights to develop and implement such principles into practice in a large industrial setting.

The webinar helped me to understand how these insights shape the research on developing principles and tools to make the A.I. principle of reliability and safety a reality, in particular when it comes to highlighting an ecosystem of open-source tools designed to accelerate machine learning (ML) life cycle, identifying and mitigating failures in a more timely, systematic, and rigorous manner.

These open-source tool-development efforts are motivated by the observation that aggregate metrics are insufficient for evaluating A.I. reliability; we need more profound insights into detailed model performance.

This ecosystem can represent an excellent opportunity for A.I. engineers and developers to glimpse into a long-term vision for integrated responsible A.I. tools that cover the entire A.I. life cycle.

What more is in for you in the webinar?

If you decide to attend the webinar, I will have the opportunity to learn about Error Analysis, a tool for identifying and diagnosing failure modes in an ML model. The diagnosis is supported by either interactive data explorations or model explanations based on InterpretML’s interpretability techniques.

Also, the webinar presents BackwardCompatibilityML — a tool for extending these insights to the scenario of model updates, to make informed decisions about which model for

deploying while accounting for regions in which an updated model progresses and regresses.

Conclusion

Justice, openness, sympathy, and robustness must be the four key pillars of responsible A.I. policy for all businesses.

While organizations cannot slow down, there is a need to unite around a set of fundamental principles of respect for the customer and a sustainable (and probably profitable) vision for long-term success.

I enjoyed this webinar and fully recommended it for anyone who is intended to live up to the commitment to being customer-centric and human-centric and want to commit to developing A.I. responsibly today.

One more thing…

I’ve just published some interesting ebooks on Amazon, and I’m sure that some of them may be interesting for you… have a look:

References

Disclaimer: I do not have any professional relationship or receive any reward from Microsoft to promote this webinar.