"You have to look at challenges as opportunities. It's not easy, but it's absolutely necessary."
Job Title: Senior Director and XAI lead
Years at the Company: 3.5
Education: University of California, Los Angeles, BA Psychology; Boston University, MS Project Management
Hometown: Lviv, Ukraine
Tell us about your role at Meta.
Irina Kofman: I lead the cross-company AI function at Meta that we call XAI. It’s made up of leaders from five different AI Innovation Centers (AIICs) across the company. The AIICs represent everything we do in AI that spans our products, research, infrastructure, integrity, and responsibility. The totality of what we do across our company in AI is amazing and unique, and it’s critical we continue to look at ways to build greater levels of connection and collaboration to realize the full potential of these capabilities.
We see a number of opportunities, particularly in how we take some of our most innovative AI technology and look for opportunities to create new experiences across our family of apps in the future. I’m excited at what we’re building and can’t wait to share more when we’re ready. Of course, the challenges are significant as well. We need to continue to look for ways to move faster, remove roadblocks, so we can accelerate the pace of innovation within a space that feels like it’s evolving daily.
This is a new role, both for you and for Meta. How did it come about?
IK: We have had our Fundamental AI Research Group (FAIR) for some time, and continue to have this well-established team working on cutting-edge research. In the past, we also had an applied AI research group, which would collaborate with different product groups. For instance, they’d go to the Ads team and say, “We have this great new technology that you can use in your products.” Sometimes this yielded great results, and other times it was a lot more challenging. Now AI centers are woven into the groups themselves, and teams are working even more closely and directly. At the same time, we need a process and team to drive overall objectives for AI at Meta, not just within individual areas. That’s the focus of my role.
How is the decentralized organizational structure working so far?
IK: We have made a ton of progress and have a lot to build on. We continue to drive breakthrough research, new product innovation, and infrastructure transformation powered by AI, and I see this pace accelerating in the future. We also have incredible support from Mark, Boz (Chief Technology Officer Andrew Bosworth), and the leadership team at Meta to prioritize our efforts and progress in AI. We have the opportunity to optimize how we communicate and collaborate so we can leverage our collective cross-company resources to move faster and at scale.
We’re sure you’re not the first person to find that managing across different teams can be something of a challenge.
IK: One hundred percent. You no longer have one org with a single leader making decisions; instead, there is a collective responsibility across the company to enable and support the success of AI. I think I am uniquely positioned for the role. I’ve worked with a lot of the people who ended up in these different groups, so there’s an established level of trust. It is also critical that we are transparent about priorities and solicit input from leaders so they understand that they’re an important part of the decision. I also try to bring in cross-functional partners early so everyone is part of the team from the beginning. I really believe the more we can show our AIIC teams the collective impact and capability of our work and their role in Meta’s success, the better we become at working together as a distributed team.
So it comes down to trust and transparency?
IK: Trust, transparency, and really strong execution. We move really quickly at Meta, and the pace of AI innovation is also happening quickly, so it is important to be able to deliver on commitments. Being reliable is key, especially with a virtual, dispersed team.
What’s the most interesting problem you’re currently working on?
IK: I think one of the most interesting challenges we have right now is in building and harnessing new teams in areas that can drive the most significant impact for Meta. Generative AI is one of these opportunities. We’ve created a new team in this area to better focus our strategy and create plans for harnessing our research learnings across our products and the community at large. Bridging research to products is never easy. But I am confident in our team and excited to share more in the near future on the exciting initiatives we have in flight.
Before joining Meta, you were an AI leader at Google, so you’ve been doing this a long time. What are some of the biggest changes you’ve seen?
IK: There’s been so much progress and change. One of the biggest areas is speech and translation. We’re now in a world where we can translate virtually any language accurately through the power of AI with the ability to scale across hundreds of languages, which makes technology more inclusive globally. We’re going to see this taken to the next level with areas like real-time speech-to-speech translation, which will give us the power to communicate with one another in ways we previously thought not possible.
I also look at areas like self-supervised learning, in which AI can learn by observation and go beyond the limitations of labeled datasets or the preconceptions a researcher might bring to a project.
And I also see continued breakthroughs in the capabilities we create for AI agents and assistants. Whether it’s to reduce cognitive overload or give you more personal and lifelike assistants, in the real world or the metaverse.
Being innovative — especially in new technologies like AI — really comes down to understanding, coping with, and learning from failure. You have to be wrong a lot before you are right, which can be discouraging. How do you keep people motivated and on track?
IK: We work in research. And research is all about the hypothesis: Did it work or didn’t it? So I think less in terms of “this project is a failure” than “we tried this hypothesis, saw it through the process, and it didn’t work. So what’s a better hypothesis?” It’s just another step on the path to the end of the project that gets you closer to the next thing. It’s precisely that ambiguity, the not knowing, that has led me to my best, most successful projects.
What’s the best way to learn from a setback?
IK: Don’t hide it, don’t diminish it, but don’t dwell on it. AI is a space that’s predicated on researchers and engineers making big bets. This is something Meta realized early, when it created the Fundamental AI Research (FAIR) group a decade ago. You have to look at challenges as opportunities, and let go of any lingering feelings of failure, and move on to iterate and get better. It’s not easy, and it can feel unnatural at times. But it’s absolutely necessary if you want to work in this field for any extended period of time.
You’re the rare woman leader in a field dominated almost entirely by men. Does that add to the pressure you feel to succeed?
IK: Sometimes. Before I took on this role, I made sure that I felt like I was the best person for the job, regardless of my gender. But the opportunity to be a woman in a leadership role in AI doesn’t come along very often, and there’s a responsibility to show up and do a tremendous job.
An added responsibility?
IK: There are always added responsibilities as a leader. When people ask me what I’m most proud of, I immediately talk about people I’ve supported: “This person did this; that person has been a huge success there.” That gives me so much energy. I know that at the end of my career I won’t be sitting here thinking, “I led AI at Meta and I launched these products,” but rather I’ll appreciate the success of my team members and the impact of what we were able to accomplish together.
How do you disconnect from work and these responsibilities?
IK: My dad started me playing tennis at an early age, and it’s been one of the things I’ve enjoyed and been passionate about my whole life. I love watching tennis, playing tennis, and the connections I have made with other players. It also enables me to manage work-life balance and commitments outside of work. It’s a sport that’s as much about mental discipline and focus as it is about athletic talent. When I look at what players like Serena Williams have accomplished, it transcends the sport and shows you how it can bring something special out of a person to share with the world.