0:0020:22

Operationalizing Responsible AI by Didem Un Ates

20:22 2022

Responsible AI fails when companies treat it as a side project

Even Microsoft, which takes responsible AI extremely seriously, has faced public AI failures, and financial liability is set to grow as regulation such as the EU AI Act arrives. Didem Un Ates, who leads applied data and AI strategy in Microsoft's Chief Data Office, hears the same plea from chief data officers again and again: skip the principles, where do I actually start?

Learn the four building blocks and six principles Microsoft uses to make responsible AI routine

Watch on YouTube

In short

Didem Un Ates, who leads applied strategy for data and AI in Microsoft's Chief Data Office, argues that responsible AI only works when it is operationalized: made standard, systematic and ongoing rather than discussed as principle. She walks through Microsoft's building blocks (principles, practices, tools and independent governance), its six responsible AI principles, and a call to action that treats responsible AI as a core business function like HR or finance.

“Operationalizing means we make it standard.”
“The easiest way to describe it, I think, is GDPR for AI. That is what it is.”

What you will learn

  • Distinguish talking about responsible AI from operationalizing it as a standard, systematic and ongoing practice.
  • Structure a responsible AI program around four building blocks: principles, team-specific practices, standardized tools and independent governance.
  • Apply Microsoft's six principles of fairness, reliability and safety, privacy and security, inclusiveness, transparency and accountability.
  • Build executive sponsorship, a cross-functional team and a measurable plan using OKRs or a scorecard.
  • Explain the EU AI Act to leaders as GDPR for AI and plan years of preparation.

Details

Level
Intermediate
Format
Talk
Technical depth
Non-technical
For
Executive, People manager, Senior individual contributor
Length
20:22
Published
July 2022

Questions this talk answers

Where should an organization start with responsible AI?

With principles, practices adapted to each team's realities, standardized tools and checklists, and ongoing governance overseen by an independent body such as Microsoft's Office of Responsible AI.

What are Microsoft's six responsible AI principles?

Fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability, which Microsoft has debated since around 2015 and invites others to adopt or adapt.

Why is operationalizing responsible AI so hard?

It requires a cultural change, not just impact assessments and toolkits, because many organizations still see AI as experimentation rather than an ongoing program.

How can leaders understand the EU AI Act?

Didem describes it as GDPR for AI, noting that organizations prepared at least three years for GDPR and AI may be even more complicated.

What should executives do to make responsible AI real?

Back it from the top, treat it as an essential business function like HR or finance, form a cross-functional team, set a plan with goals and metrics, and use shared toolkits.

Topics

Related talks

Browse the library