DataHack Radio #10: Jeannette Wing on Building Trustworthy AI with Strong Foundations in Computer Science

The meteoric rise of data science and artificial intelligence is reshaping virtually every domain, from healthcare and finance to transportation and entertainment. Consider these astonishing statistics:

• The global AI market is forecasted to grow from $387 billion in 2022 to over $1.3 trillion by 2029, a 20% annual growth rate. (Source)
• Over 64% of businesses report investing in Big Data and AI initiatives. (Source)
• Machine learning patents grew at a 34% CAGR between 2013 and 2017, the fastest growth rate among all patent categories. (Source)

However, the theoretical foundations that enable these cutting-edge technologies are rooted in decades of pioneering work in computer science. One visionary leader who has not only built these foundations but connected them to the frontiers of data science and AI is Dr. Jeannette Wing.

Currently the Avanessians Director of Columbia University‘s Data Science Institute and a Professor of Computer Science, Dr. Wing has over 40 years of experience advancing the field. She has held leadership roles at Carnegie Mellon University, Microsoft Research, and the National Science Foundation. Her research spans areas including formal methods, trustworthy AI, and the foundations of data science.

In this exclusive DataHack Radio podcast, Dr. Wing shares her unique perspective on the vital role that computer science plays in the data science revolution. She elucidates how concepts from theoretical computer science, such as formal methods and complexity theory, are powering the latest breakthroughs in machine learning and AI. By taking a principled approach grounded in strong mathematical foundations, Dr. Wing argues that we can build data science and AI systems that are more robust, explainable, and aligned with human values.

Formal Methods: A Key to Trustworthy AI

One of Dr. Wing‘s core research areas is to use formal methods—rigorous mathematical techniques for specifying and verifying system properties—to help build more trustworthy AI and machine learning systems. While formal methods have traditionally been applied to software and hardware systems, Dr. Wing sees them as critical for modern data-driven systems as well.

"As we see the ubiquity of AI and machine learning across all sectors of our lives, it‘s becoming increasingly important that these systems are safe, secure, and have properties like fairness and robustness," Dr. Wing explains. "Formal methods give us a way to precisely specify and verify these properties."

For example, Dr. Wing cites recent work using formal methods to verify the fairness of machine learning algorithms, such as ensuring that a facial recognition model performs equally well across different subgroups. By encoding fairness properties in mathematical logic and using formal verification techniques like model checking, we can provably guarantee that an AI system satisfies certain fairness criteria.

Other important properties that formal methods can help ensure include:

• Safety: verifying that an AI system will not produce harmful outputs or take dangerous actions
• Security: proving that an AI system is resilient to adversarial attacks and manipulation
• Robustness: ensuring that an AI system performs consistently well even with noisy or unexpected inputs
• Interpretability: enabling humans to understand the reasoning behind an AI system‘s predictions or decisions

"With formal methods, we can lift the level of abstraction and express high-level properties that we want these AI and machine learning systems to have," says Dr. Wing. "Then we can use mathematical and logical techniques to verify that those properties are satisfied."

Data Science as a Confluence of Fields

In addition to her research on trustworthy AI, Dr. Wing has been a thought leader on the foundations and future directions of data science as a field. She sees data science as a unique confluence of multiple disciplines, including computer science, statistics, mathematics, and domain expertise.

"Data science is really at the intersection of all these different fields," Dr. Wing explains. "From computer science, it draws on areas like algorithms, databases, and distributed systems to deal with large-scale data processing and computation. From statistics and mathematics, it leverages probability theory, optimization, and signal processing to extract insights from noisy and high-dimensional data."

However, Dr. Wing emphasizes that data science is more than just the sum of its parts. "What makes data science unique is that it‘s fundamentally driven by real-world problems and domain-specific questions," she says. "Data science is about using all these computational and statistical tools to actually answer questions and drive decision-making in a particular domain, whether it‘s healthcare, finance, climate science, or any other field."

This interdisciplinary nature of data science is reflected in Dr. Wing‘s leadership of Columbia‘s Data Science Institute, which brings together over 200 affiliated faculty from 12 different schools across the university. "We have collaborations with the medical center, the business school, the journalism school, and of course the engineering school," Dr. Wing shares. "The institute is really a hub for data science research and education across the entire university."

The Global Future of Data Science and AI

Looking to the future, Dr. Wing sees the impact of data science and AI only continuing to grow and expand across the globe. She highlights the rapid advancement of fields like quantum computing and computational biology as two areas where data science may have transformative impacts in the coming years.

"Quantum machine learning is an exciting area that could lead to exponential speedups for certain types of data science problems," Dr. Wing explains. "And in computational biology, data science is enabling us to analyze massive genomic and proteomic datasets to gain new insights into the mechanisms of diseases and accelerate drug discovery."

Dr. Wing also emphasizes the increasingly global nature of data science research and innovation. "Data science and AI are advancing rapidly not just in the US but around the world," she notes. "In my time at Microsoft Research, I saw firsthand the incredible work happening in places like China, India, and the UK."

She cites Microsoft Research India in Bangalore as a particular hub of cutting-edge data science research, with projects spanning areas like technology for emerging markets, computational agriculture, and societal resilience. "There‘s a real vibrancy and energy in the data science research community in India," Dr. Wing describes. "They‘re not only advancing the state-of-the-art in machine learning and AI, but also focusing on applications to local challenges and opportunities."

Conclusion

As data science and AI technologies become more powerful and pervasive, Dr. Jeannette Wing offers a vital perspective on the importance of strong foundations in computer science. Through her pioneering research and leadership, Dr. Wing is connecting the core concepts and techniques of theoretical computer science to the frontiers of data-driven discovery and decision-making.

By grounding data science and AI in rigorous approaches like formal methods, Dr. Wing argues that we can build systems that are not only more capable but more reliable, understandable, and aligned with human values. At the same time, she paints an exciting picture of the future of data science as an expansive, global endeavor that intersects with emerging areas like quantum computing and computational biology.

"It‘s an incredibly exciting time for data science and AI," Dr. Wing concludes. "We have an opportunity and a responsibility to shape these technologies in ways that benefit science, society, and the human condition."

To dive deeper into Dr. Wing‘s vision and insights, be sure to listen to the full DataHack Radio podcast episode. And keep an eye out for new breakthroughs and initiatives from her team at Columbia‘s Data Science Institute as they advance the foundations and frontiers of data science and AI.