Maya Ajmera, President & CEO of Society for Science & the Public and Publisher ofScience News, sat down to chat with Lester Mackey, a statistical machine learning researcher at Microsoft Research Labs New England. Mackey is an alumnus of the Science Talent Search (STS) and the International Science and Engineering Fair (ISEF), both science competitions of the Society. We are thrilled to share an edited summary of the conversation. 

You’re an STS 2003 and ISEF 2003 alum. How did the competitions affect your life?

When I was a collegefreshman at Princeton, an Intel researcher reached out to me and asked if Iwanted to intern in their Strategic CAD Labs. The lab was made up entirely ofPh.D.s and hadn’t taken on interns before. The researcher knew of me because ofmy participation in ISEF and STS, and if it hadn’t been for my science fairparticipation, I wouldn’t have had the opportunity to work there.

Long story short, I lovedthe experience. I liked the freedom that I had, the opportunity that I had toproductively deploy creativity at every turn. And I came away from thatexperience determined to get a Ph.D. What’s more, my Intel mentor knew MariaKlawe, who was dean of Princeton University’s School of Engineering and AppliedScience at the time, and recommended that she recruit me for a researchproject. That kicked off my research career.

Do you have any memorable experiences from the competitions to share?

My ISEF project wasawarded an Operation Cherry Blossom Award, which included an all-expense-paidtrip to Japan. This took me on an incredible adventure: We went to Tokyo andYokohama and the ancient capitals of Nara and Kyoto. Part of the trip involvedmeeting a Japanese princess.

You taught at Stanford before moving to Microsoft. Can you describe how being at Microsoft has been different than a purely academic environment and how that’s helped you in your career?

In many ways my lab,Microsoft Research Labs New England (MSR), is very much like a university. Theresearchers here work on whatever they want. We publish everything. We’reevaluated based on our contributions to our academic communities and to theworld.

The main difference is inthe extra degree of freedom that we have at MSR. If you want to spend 100percent of your time doing research, you can do that. If you want to teachcourses at neighboring universities, you can do that too. You have the freedomto choose how you want to spend your day, and I think that freedom is veryvaluable.

I’ve also noticed thatthe researchers here tend to be very hands-on with their projects and verycollaborative. I find myself working not just with my students, interns andpostdocs, but also with my talented and experienced labmates. That’s been a bigpositive for me. Our lab is somewhat unique in that it was created as aresearch lab for both computer scientists and social scientists. Some of my colleaguesare scholars in economics, communication or anthropology. The lab fits on asingle floor, and that proximity breeds collaboration. I find myself working onproblems that I hadn’t even considered before coming to MSR.

How do you recommend students start studying and getting involved in machine learning?

I tell all students totry a data science competition. My first encounter with machine learning wasthrough a competition that Netflix ran when I was a senior in college. Netflixwanted to improve its movie recommendation system, so they released a datasetof 100 million ratings that users had given to various movies. Competitors werechallenged to predict how the users would rate other movies in the future. Myphilosophy is that these public competitions provide a great sandbox for peoplewho are just starting out in machine learning because you get to work with realdata on a real problem that someone really cares about. By the end, you’llunderstand both the methods and the problem, and you’ll have fun doing it.

There’s been a lot of talk about artificial intelligence and its influence on humankind. Why do you think students or the public should care about AI?

I think we have to beaware of these technologies so that we can hold them accountable to ourstandards of fairness and safety. AI is becoming much more pervasive, and it’sincreasingly being incorporated into technologies that impact our everydaylives: self-driving cars, résumé-screening tools and algorithmicrisk-assessment tools that inform bail-release and criminal-sentencingdecisions.

I also think that AIholds the potential to help us address some of our biggest challenges likepoverty, food scarcity and climate change. What I love most about my field isthat these tools have the potential to do real good. I think that’s somethingthat will excite many students and the public more generally.

Using AI to solve issues like poverty is interesting to think about. I would love to hear more about how this technology can be employed to solve these types of problems.

Take the example ofGiveDirectly, a nonprofit that gives unconditional cash transfers to thepoorest people in the poorest communities. They’re finding that this leads tosustained increases in assets. However, the on-the-ground process theorganization goes through to identify transfer candidates is quite laboriousand expensive.

So they’ve been workingwith experts in machine learning, statistics and data science to automate moreof that process. Early work transforms satellite images into predicted povertyheat maps to guide the search of field-workers, and I think we’ve justscratched the surface of what is possible.

What do you feel are the most interesting problems that could be addressed within your field of research?

I’d love to see the fielddirect more of its attention and resources to social problems like poverty,hunger and homelessness. There are many open questions in this space. Whatspecific problems could actually benefit from machine learning intervention?How can machine learners work with experts and policy makers to actually affectmeaningful change? How do we incentivize our talented students and professionalmachine learners to work on these problems?

A second, different sortof challenge is responsible deployment. We see that AI is being used already toinform important decisions in society, such as screening résumés or determiningwhen people should be released on bail. How do we ensure that those decisionsare fair and reflect our societal values? This is an increasingly active areaof research in the field.

As a person of color in machine learning, what are your thoughts on bringing more young people of color into this field?

There have been somedevelopments in this direction that I’m particularly excited about. There’s a“Black in AI” movement now. It’s bringing people of African ancestry togetherin this field. Although we might be sparse and distributed, we have a bigpresence. It’s been excellent for networking and for encouraging younger peopleto get involved in machine learning and stay involved. You can learn more aboutit at blackinai.github.io or by searching online for Black in AI.

What books are you reading now? And what books inspired you when you were younger?

I just finished readingAstrophysicsfor People in a Hurry by Neil deGrasse Tyson, and now I’m in the middle ofAmericanNations by Colin Woodard.

When I was younger, my threefavorite books were Brian Greene’sThe Fabric of the Cosmos, MattRidley’sThe Red Queen and Neil Gaiman’sAmerican Gods.

The world faces so many challenges today. What keeps you up at night?

I would say poverty. Ican’t comprehend how there can be so much wealth in my field, my community andthis country, and yet half a million people in the United States are homelesson any given night. One in nine people are malnourished in the world. That justdoesn’t make sense to me.