CGS Employer Roundtable member, Dr. Lisa Amini, is at the cutting edge of advanced machine learning and artificial intelligence (AI). She currently serves as director of IBM Research Cambridge, home to the MIT-IBM Watson AI Lab and IBM’s AI Horizons Network.
In her more than 25 years at IBM Research, Dr. Amini has held a number of positions, including her current role as director of AI scaling and automation, and former roles as the director of knowledge & reasoning research in the Cognitive Computing group at IBM’s TJ Watson Research Center, founding director of IBM Research Ireland, and the first woman lab director for an IBM Research Global Lab. Dr. Amini earned her PhD in computer science from Columbia University.
We sat down with Lisa to talk about the role of graduate education in preparing researchers to solve the problems of the future.
As an employer, what do you see as the value of hiring people with advanced degrees?
We value advanced degrees, because they help to build deep expertise in relevant areas such as Artificial Intelligence or mathematics, and to tackle problems leveraging scientific methodologies. An IBM researcher strives to advance science or to envision the future of technology, and then work towards creating that future. Accomplishing this requires two things. One is deep expertise in their chosen field, along with sufficient breadth in related areas, for a broader context. That depth and breadth of expertise requires advanced courses and research project experiences available in graduate school.
The second is that working at IBM Research requires leveraging scientific principles and methods, to bring rigor and theoretical underpinnings to our research. Working at the forefront of science and technology, requires tackling problems that are not initially well-posed. Much of a researcher’s job is in formulating challenges as more well-defined problems, which can then be tackled with principled approaches. These skills are usually developed and refined through graduate work. When we’re hiring people with PhDs, we look at their coursework and also their research results. We want to hire people who can demonstrate their ability to formulate impactful problems, generate new ideas, and communicate their research to peers in the scientific community, for example through papers and presentations in top-tier conferences and journals.
Your company does a lot of work to address complex societal challenges. Tell us about one or two of those initiatives and your role in them?
I’ll start with one of the biggest projects I took on personally to tackle complex societal challenges: the Smarter Cities Initiative. In 2003, I had started researching capabilities to analyze streams of data and extract patterns and insights as the data streams were being generated. We were inventing scalable stream mining systems and machine learning algorithms to handle what we saw as a growing tide of data that would be continuously generated from large numbers of sensors and devices.
In 2009-2010, I was asked to start a new research lab in Ireland to extend and leverage these ideas and technologies for building smarter cities and utilities, such as energy, water and transportation. This was an exciting initiative because at that time, much of the machine learning research on very large data leveraged synthetic data. Large, real-world data sets were not widely available in academic settings. It was also exciting because there were huge efficiencies and societal benefits to be gained by improving these systems. This enabled to engage with organizations seeking to deploy smarter transportation systems, transportation grids, water networks, and large energy grids.
We were able to take our expertise in stream processing, machine learning, optimization and control theory, and apply it to real-world challenges. Our goal was to help people run a more efficient city by providing better technology to those managing city operations or utilities. It was also exciting because we were building a new research lab from the ground up. It was a compelling mission, and we were one of the few labs getting that magnitude of data and access to address real problems.
One of the more recent projects you’ve probably seen us talking about is our work in AI that drives data-driven discovery using generative AI approaches. This is one of the areas we are tackling as part of our AI Horizons Network (AIHN) of university partners. You often see this research in terms of the future of materials, molecular discovery, or medicine. This is such a large and challenging space with huge potential for societal impact, it is important for us to create communities of discovery to enable better science, better collaboration and sharing, and better ability to reproduce and build upon research results across those communities of discovery.
This also ties back to the initiative I lead within IBM Research, the automation of AI. By automating the processes of AI model creation and operations, such as, feature engineering, machine learning (ML) model building, and ongoing ML lifecycle management, we hope to bring AI systems with better repeatability, transparency, fairness, and scalability to these problems. This is because the AI automation algorithms seek to bring best of breed algorithms and techniques to ML tasks and pipelines. And I hope that these capabilities will be one of the foundations for these communities of discovery we see emerging.
What advice do you have for students, particularly women in STEM thinking about graduate school?
If you want to get an advanced degree, do it in on a topic you’re deeply passionate about. The work is not easy, and there will be times when you’re struggling to meet the demands. So, you really need that passion to keep going. If you finish your undergraduate degree and aren’t sure which fields or problems you should bring your passions to, consider taking time to work in the industry and then decide when you are ready to pursue an advanced degree. I actually worked in industry for a while before returning to grad school for my doctoral degree.
I would also say not to worry if your path isn’t linear. You may start off in a particular area and find something else you enjoy more. It’s not like you lose that knowledge or experience. I’ve heard people say things like, “I’m already a year into my program, I can’t change tracks, because I will lose the time I’ve already invested.” In my opinion, graduate school is the time to experiment and learn, and every new thing you learn helps you to build your own personal body of knowledge. You will be faced with other pivots throughout your career, so all these little shards of evidence help you to reason better and to find your space.
The last point is that regardless of your field of study, you should learn data-driven approaches, techniques, and tools. We see more people pursuing fields such as finance, economics, social science and others, while also learning machine learning and data science methods and tools. Using grad school to learn these techniques early will enable you to bring data-driven grounding to your research in any domain, throughout your career.