reMARS revisited: Frontiers of fair and accessible AI
Prem Natarajan, Alexa AI vice president, and Michael Kearns, an Amazon Scholar, discuss fairness, accountability, transparency, and ethics topics applied to machine learning, automation, robotics, and space themes.
In June 2022, Amazon re:MARS, the company’s in-person event that explores advancements and practical applications within machine learning, automation, robotics, and space (MARS), took place in Las Vegas. The event brought together thought leaders and technical experts building the future of artificial intelligence and machine learning, and included keynote talks, innovation spotlights, and a series of breakout-session talks.
Now, in our re:MARS revisited series, Amazon Science is taking a look back at some of the keynotes, and breakout session talks from the conference. We've asked presenters three questions about their talks, and provide the full video of their presentation.
On June 23, Prem Natarajan, Alexa AI vice president, and Michael Kearns, an Amazon Scholar who is a professor and National Center Chair in the Department of Computer and Information Science at the University of Pennsylvania, presented their talk, "Frontiers of fair and accessible AI". Their talk focused on their perspectives on fairness, accountability, transparency, and ethics (FATE) topics applied to machine learning, automation, robotics, and space themes.
What was the central theme of your presentation?
Our vision at Amazon is to make AI useful for every user worldwide in their everyday lives by advancing ambient intelligence. This talk highlights the progress and opportunities with building AI services that deliver on that vision, making the benefits of AI broadly accessible to all users. We adopt a multidisciplinary approach that is (1) rooted in inclusive AI design; (2) advances and builds on the state-of-the-art in AI/ML; and (3) brings together the best minds from academia and industry to address the hardest challenges in AI. We also describe how we build Alexa experiences that are fair, inclusive, and accessible for our global and diverse customer base, and share recent AI/ML advances accomplished by our scientists and engineers to detect and address performance variations in automatic speech recognition and natural language understanding production models at scale. We present case studies on adaptive listening and Alexa Together, and provide a forward-looking perspective with mobility-centric devices such as Astro.
In what applications do you expect this work to have the biggest impact?
Our methodology is based on a generalizable AI framework which is broadly applicable to several different use cases including conversational AI, explainable or interpretable AI, and creative or content-generation applications.
What are the key points you hope audiences take away from your talk?
- Developing inclusive, fair, and broadly accessible AI applications requires a full lifecycle perspective that spans design, technical invention and innovation, and includes diverse perspectives at every stage of the development process.
- Generalizable AI frameworks such as foundation models can be harnessed to make AI applications more useful for all customers, and thereby make the benefits of ambient intelligence broadly accessible to everyone.
- Multi-sector partnerships, especially between academia and industry are crucial for addressing the hardest challenges in AI such as reasoning and teachability. Amazon leads the way through its investments in entrepreneurs from underrepresented backgrounds, collaborative sponsorship of academic research with the National Science Foundation Collaboration, and targeted programs (such as Amazon SURE in collaboration with Columbia, USC Viterbi, and Georgia Tech) that provide career-shaping opportunities to students from diverse backgrounds.