Dr. Gillian Hadfield - Explainable vs. Justifiable AI

A broad thinker from an unusual background, Dr. Gillian Hadfield shares a different take on building these models from the general norm, as well as how to incorporate transparency into justifiable systems, and the hypothesis of building a system where decisions are attached back to a person responsible. We also talk about the need for safe, consistent, and up-to-date regulatory structures, and the effects of not having this, before closing with some powerful advice around the work we have to do going forward in this sector! We hope you can join us for this hugely insightful conversation.

Key Points From This Episode:

• Introducing Dr. Gillian Hadfield and what drew her to the space of law and globalization.
• How the challenges of AI align with the challenges of economics.
• The need for people in social sciences and humanities to engage in design and building.
• The objective of the Schwartz Reisman Institute for Technology and Society.
• Defining AI governance to address the alignment problem.
• Comparing AI with conventional programming and the difficulties with test sets.
• The difference between AI explainability and justifiability.
• Talking about the GDPR and what they are really looking for.
• A legal analogy on incorporating transparency into justifiable systems.
• Discussing the chicken-and-egg confusion that regulators are feeling.
• Why we haven't seen the growth of AI we would expect.
• How regulatory regimes haven't kept up with the speed of globalization and digitization.
• The balance of having the right kind of regulation.
• A walk through the current landscape of AI regulation.
• What regulatory technologies look like.
• The focus on fairness and algorithmic bias and AI's capacity in all domains.
• Dr. Hadfield's advice for people who are looking for AI integration in their practices.

Tweetables:

“There's no one solution to how you align AI.” — @ghadfield [0:10:19]

“We have the alignment problem everywhere. How do you get a corporation to do what you want it to do, how do you get governments to do what you want them to do?” — @ghadfield [0:23:52]

“AI is a general-purpose technology, it's a way of solving problems, it's a way of coming up with new ideas. It's going to be everywhere. I prefer to think of the regulatory challenge as, how is AI changing your capacity to achieve your regulatory goals, in any domain?” — @ghadfield [0:37:41]

“We need way more people who are not engineers, deeply engaged in the process of building our systems.” — @ghadfield [0:42:13]

Links Mentioned in Today’s Episode:

Gillian Hadfield on LinkedIn

Gillian Hadfield on Twitter

The Vector Institute

Schwartz Reisman Institute for Technology and Society

GDPR

Mila