Behind the Research with Giuseppe Paolo

An inside look into the work and expertise from research scientist, Giuseppe Paolo

Welcome to the fourth edition of our Behind the Research blog series, featuring Cognizant AI Lab researchers. This series offers an inside look at some of their work and what inspires them, what they’re interested in, and how it shapes the technology that is becoming part of our daily lives. Today, we are interviewing Giuseppe Paolo, Associate Director and Research Scientidst at our AI Lab.

Firstly, how did you first get interested in AI and research more generally?

• It started during my bachelor's degree, when I discovered robotics. Studying robotics led me deeper into AI, and I immediately thought: this is really cool. I knew I wanted to work in AI, so I pivoted toward research. It was around 2016, when AI was still fairly new and rapidly developing. I got really into reinforcement learning and became fascinated by how it could be applied to AI systems.

What are the main problems you are currently working on, and what makes them meaningful or exciting to you?

• I'm mainly focused on two things right now. The first is the study of multi-agent systems  when you have multiple AI agents working together. What do they actually create? Do they do whatever they want, or do they respect the goals you've set? How do they develop new behaviors, and when do they go off the rails? As we move into a future where more of these agents will run more freely, understanding what they do becomes critically important.

• The second project I'm excited about is around whale-ship collision prevention, which is the idea of using AI to help shipping companies avoid crashes between ships and whales. We're using AI for good here, to genuinely improve society. Right now, we're working through the literature to understand what's already out there and figure out how to make this work in practice.

How do you decide which research questions are worth pursuing and exploring in depth?

• I usually start with something that genuinely interests me, then read the literature around it. From there, I start talking to people and getting a sense of whether the idea is repetitive or whether there's something worth pursuing. Eventually, I just start trying things like start coding or start experimenting. If interesting results come out, I keep going. If nothing interesting emerges, I move on.

Can you share a project you’re especially proud of and what you learned from it?

• The project I'm most excited about right now is TerraLingua,  a simulation where multiple agents have to interact with each other, and we study what they do. What I've learned is that AI agents tend to be surprisingly collaborative and friendly, even when they're pitted against each other. They try to be helpful. That's interesting, because so much of the popular narrative is about an "AI apocalypse," but these behaviors run completely counter to that.

How do you stay up to date with such a fast-moving field? And are there any areas of AI that you think are underrated or not getting enough attention right now?

• For staying current, I surf the web, Twitter, Reddit, and LinkedIn for the latest news. I'm also subscribed to email groups where I receive the latest papers from arXiv. And I use tools like ChatGPT and Claude to help with literature research.

• In terms of underrated areas, I think multi-agent research and embodied AI — AI applied to robotics or physical systems — deserve a lot more attention than they currently get.
  

Looking back, is there anything you would have done differently in your path toward AI research? What advice would you give to someone who is interested in getting into AI research?

• I wish I'd pushed myself toward more applied problems earlier. Research doesn't always connect to real-world applications as there are a lot of toy problems and toy environments and I think working on real-world challenges is where the most meaningful impact happens.

• My advice to anyone getting into AI research: don't follow the huge trends on social media. It's very easy to get FOMO and feel like you're falling behind. Instead, find the subfield that genuinely interests you and go deep on that.

Your AI agent has a dashboard of “human stats” about you. What three silly metrics is it tracking?

• Number of cups of coffee consumed throughout the day

• Pizzas eaten per week

• How much time I spend talking about motorbikes