Teaching fairness to machines

By Mike Schroepfer
March 10, 2021

Almost everywhere you look these days, artificial intelligence is making dramatic progress, from teaching machines to translate languages to understanding the visual world. And one great part of my job is getting to speak with the researchers who are advancing the state of the art in all these areas.

Recommended Reading

But I believe one of the most important fields of AI is the one that will help all the others actually deliver on their promise to the world: fairness. Designing AI to treat people fairly and equally is an extremely hard problem — it begins with fundamental questions about what fairness means in each context, and goes all the way to the technical challenge of turning those definitions into code. Today we shared some more detail about how we're approaching that technical challenge, which you can read about here

I recently spoke with someone doing groundbreaking work on this: Rachad Alao, who leads Facebook’s Responsible AI efforts. Among many other things, Rachad and his team have worked to advance the state of the art in privacy-preserving machine learning. We talked about that, and more, in this video:

The coming decade of technological innovation, from virtual reality to self-driving cars,  will be enabled in large part by artificial intelligence. In every single field it touches, it will be mission critical for AI to treat people fairly and equally. That’s why Rachad’s work is so important, and something we can all learn from.