Meta is announcing the Artificial Intelligence Learning Alliance (AILA), an initiative to strengthen diversity and increase equity in the field of artificial intelligence.

Artificial intelligence (AI) offers important benefits for people in areas such as transportation, commerce, and beyond. While AI is constantly improving at understanding the dynamics of the world we live in, there is much more work to do. We’re firm believers that it will progress faster with early involvement from experts with diverse perspectives. If the researchers, developers, and engineers advancing AI come from a limited variety of backgrounds, the technology’s point of view will be similarly limited.

Addressing this starts in classrooms. That’s why Meta, together with Georgia Tech, created the AI Learning Alliance (AILA), an expansion of the co-teaching program that we piloted in the Fall of 2020. Meta AI researchers both developed the curriculum for and guest-lectured in the deep learning course that Georgia Tech assistant professor Zsolt Kira has taught for the past two years as part of the university’s Online Master’s of Science in Computer Science program (OMSCS). The goal was to help students experience real-world examples and learn valuable techniques used in deploying and scaling algorithms.

The statistics above were taken from the 50th annual Taulbee Survey. (https://cra.org/wp-content/uploads/2021/05/2020-CRA-Taulbee-Survey.pdf)

The original plan was to teach the course on campus in Atlanta, but due to the pandemic, all course instruction was moved to a virtual model after only three in-person classes. In the end, more than 1,600 students took the course in its inaugural year.

Now, through our online learning platform Meta Blueprint, we’re poised to teach thousands more by opening the AILA Education Hub up to everyone — educators, students, researchers, and hobbyists alike. We are also developing two new courses to complement AILA’s deep learning program: Introduction to AI and Natural Language Processing. Both will be added to the hub in 2022 and are geared toward those who have a computer science background. Students will also gain important and relevant real-world experience by working with members of Meta AI teams on active projects from the open-source community like PyTorch, the Hateful Memes Challenge, fastMRI, and AI Habitat. These new courses will be added to the hub in 2022. By eliminating barriers to learning and increasing access to opportunity we hope to help increase the number of diverse students studying AI and open new pathways to careers in the AI field for people from underrepresented groups.

“Many people who want to work in AI cannot access the resources they need to get there,” said Charles Isbell, the Dean and John P. Imlay, Jr., Chair of Computing at Georgia Tech. “By moving AI instruction online, we can reach more people from a wider range of backgrounds than ever before. This is not only a great opportunity for learners, but also for the field as a whole, which needs a diverse set of voices if it is to responsibly serve a diverse set of communities.”

We believe deeply in this model, so we took the additional step of creating a consortium of professors who see value in the curriculum we’re developing and are open to teaching some or all of the material at their universities in various capacities including a mix of both online and in-person offerings. These professors teach at Historically Black Colleges and Universities (HBCUs), Hispanic-serving Institutions (HSIs), and Asian American/Native American Pacific Islander-serving Institutions (AANAPISIs):

  • Georgia Tech University

  • Florida Agricultural and Mechanical University (through their foundation)

  • Morgan State University (through their foundation)

  • Alabama A&M University (through their foundation)

  • University of Central Florida (through their research foundation)

  • Georgia State University (through their research foundation)

  • Florida International University (through their foundation)

  • University of California Irvine (through their foundation)

  • North Carolina A&T State University

To further support these universities, Meta partnered with Dr. Kira at Georgia Tech to create a series of webinars which will be hosted on the AILA Education Hub and are intended to help guide professors on how to teach the course content. We’ll host, record, and share Q&A sessions and work through modules and exercises related to the curriculum.

Professors at universities around the world know first hand how challenging it is to keep graduate-level CS students up to date on the latest deep learning techniques being applied today by scientists and researchers.

“We have several courses in which we plan to explore system integration, including our classes on Introductory ML and Advanced ML (Deep Learning), as well as Natural Language Processing,” says Mark A. Finlayson, an Eminent Scholar Chaired Associate Professor of Computer Science in the Knight Foundation School of Computing and Information Sciences at Florida International University. “These fields move very quickly and it makes sense to partner with other academics and organizations to reduce the effort required to keep our curriculum up to date with high quality materials.”

Teaming up with Georgia Tech to create AILA was a deeply thought out partnership that took a lot of effort. We were looking to collaborate with a university known for producing large numbers of graduates from diverse backgrounds, and OMSCS is one of the country’s leading programs in doing just that.

The school’s online CS Master’s program is the largest of its kind in the United States and is one of the country’s most racial and ethnic diverse universities. Its laudable success at enrolling members of underrepresented groups into graduate-level CS programs makes it the perfect institution for us to team up with in pursuit of our longstanding goal of increasing diversity in the field of AI.

This is crucial given that the number of people from underrepresented groups earning advanced degrees in computer science is consistently low. In 2020, according to the Computing Research Association’s Taulbee Survey, just 3.8% of CS doctoral graduates were Black, Hispanic, or Native American.

The statistics above were taken from the 50th annual Taulbee Survey. (https://cra.org/wp-content/uploads/2021/05/2020-CRA-Taulbee-Survey.pdf)

 

If we don’t grow that number, it’s unlikely that the algorithms used to detect diseases, improve transportation, or even help connect people, will adequately reflect the diverse audience they are intended to serve. As the World Economic Forum puts it, by “addressing diversity gaps today, tech leaders can mitigate bias in the systems built for tomorrow.” Working alongside schools that enroll and support large numbers of students from underrepresented groups, we’re aiming to do just that. Collectively, we’re putting the curriculum in the hands of educators and truly anyone who wants to learn, creating a gateway to this content that requires no upfront investment. All they need is the time for taking the course, an appetite for learning, and if they choose, to avail themselves of an opportunity to pursue a career in AI. This is just one piece of our ongoing commitment to manifesting this reality.

“Morgan State promotes unbiased and trustworthy AI and AILA provides cutting edge curricula that enables students to learn the knowledge and skills from top industry experts and closes the gap between academia and industry,” says Dr. Paul Wang, Director of the Trustworthy AI Lab and Professor and Chair of Computer Science at Morgan State University. “The collaboration between Morgan Computer Science and Meta transforms learning to our mostly underrepresented students and prepares them to enter the workforce with competence and confidence.”

A curriculum built for the real world

Our semester-long deep learning course teaches the fundamentals of neural networks and applications like computer vision and language understanding that underpin some of today’s most-important industrial uses of AI. The course bridges AI theory and application and covers deploying and scaling algorithms in the real world. This approach has earned high marks from those who took the course.

Please click here to register for the course or to learn more about it.

 

By Denise Hernandez, Meta AI Program Manager
Source Facebook AI Research

Previous System Cards, A New Resource For Understanding How AI Systems Work
Next Project CAIRaoke: Building The Assistants Of The Future With Breakthroughs In Conversational AI