New LTI Director To Usher in 'Responsible Thinking' at CMU

Mona Diab Has Eyes Set on Changing How To Think About AI

Mona Diab understands what is at stake.

As a research scientist at two of the largest technology companies on the planet, Diab saw the impact innovations had as they spread across the globe. And with artificial intelligence poised to usher in the greatest technological leap since the internet, Diab wants to train, teach and prepare students, researchers, scientists and communities to think responsibly about these new tools.

"We're living in a world of proliferating AI and generative AI. There is so much at stake," Diab said. "We are at an inflection point for this technology and the discipline as a whole. It is a critical time to take into account its impact and sustainability."

Diab takes the reins of Carnegie Mellon University's Language Technologies Institute (LTI) this month. Part of CMU's School of Computer Science, the institute is home to cutting-edge research into the large language models powering the recent wave of generative AI systems and has a history of developing systems that have changed how computers understand and interact with humanity. 

For Diab, AI researchers, scientists and students must consider issues ranging from security, privacy and accountability to diversity, equity and inclusion. They need to ask questions about climate and culture alongside questions about ones and zeros. Diab coined this the "responsible thinking" approach, and she wants the LTI to embrace it as fully as critical and computational thinking.

"Responsible thinking needs to become core to our work. It should shape the future of language technologies and AI at large" Diab said.

Diab joins CMU from Meta, where she was the lead responsible AI scientist. Before Meta, Diab worked as a principal scientist at Amazon Web Services and was a professor at George Washington University, where she founded the CARE4Lang lab. She is a globally renowned expert on Arabic natural language processing (NLP), multilingual processing and computational social sciences. She earned a doctorate in computational linguistics from the University of Maryland and a master's degree in computer science from George Washington University.

She majored in computer science and Egyptology with undergraduate degrees from The American University and Helwan University, both in Cairo. Diab blends both of those majors in a research thrust using NLP technology to build tools to read and reconstruct hieroglyphics in hopes of opening the field to more people and gaining new insights into history.

CMU's trailblazing research and top AI and computer science programs and talent drew Diab to the university. She said there is no better place to affect the development and use of the next generation of technology than the School of Computer Science.

Diab believes there is a gap in the current workforce, as it doesn't cater to what the current technology needs. To fill that gap, she wants to instigate a strong student-centric paradigm shift at CMU, to sustain excellence in both research and pedagogy. She aspires to train the best possible talent and transform the LTI into a leader in responsible thinking. With top faculty training top students who go on to secure top positions in industry and academia, the institute is positioned to snowball a responsible posture toward AI. 

"Mona Diab's relentless pursuit of responsible AI in both academia and industry will strengthen the work of the Language Technologies Institute, School of Computer Science and Carnegie Mellon University as we seek to empower future generations of AI talent and develop a new wave of transformative technologies in an ethical, fair and sustainable manner," said Martial Hebert, dean of the School of Computer Science. "We are thrilled to have Mona's expertise and passion at CMU and excited to work with her on her vision."

Diab wants to proactively build the responsible AI framework into technology from the onset to avoid potential cultural harm as it spreads across the globe. She saw these harms as social media platforms spread to other countries, where Western-built platforms couldn't accommodate the nuances of different cultures.

Her work at Meta changed how the company handled Arabic, challenging the company to abandon the previous notion that it was a monolithic language and instead acknowledge the rich and diverse variants spoken around the world. That broader understanding helped Meta better grasp the language, leading to better translations, better identification of hate speech and a better user experience.

"I'm Muslim and I'm Arab, and I'm concerned that a lot of this technology is going to be adopted blindly," Diab said. "We must embed the technology with the right frameworks and build systems that are culturally aware and culturally responsible."

Diab advocates for computer scientists to work alongside social scientists. Anthropologists, sociologists, ethicists and others can help AI engineers think about problems from culturally different points of view and tune those systems to improve them. But this is not a one-way street. Diab believes that the relevant social sciences can also benefit from working with computer scientists.

Developing AI technology must also happen with the people it intends to serve, and feedback from that population should guide evaluations.

"It's in our interest to learn from them and assess the success of our systems with them," Diab said. "Developing these strong collaborative initiatives and outreach will benefit all of humanity."

She strongly advocates for effective diversity and inclusion and welcoming everyone equally in the scientific landscape. It is critical to the sustainability of the scientific enterprise. 

Diab will be the LTI's second full-time director. The institute's founder and first director, Jaime Carbonell, died in February 2020. Professors Jamie Callan and Carolyn Rosé have served as interim directors.

The LTI, world-renowned for its pioneering work on natural language processing, machine translation, speech recognition and large language models, was founded in 1986 as the Center for Machine Translation and changed its name in 1996. The institute offers a Ph.D. in language and information technology, and master's degrees in language technologies, intelligent information systems, computational data science, and artificial intelligence and innovation. Undergraduates can complete a concentration in language technologies. The institute has 28 core faculty members, a staff of 41 including post-doctoral researchers, and 461 students.