Interview with Paul Saunders

AI Researcher

Download Original (.pdf)

What’s one major AI research trend or technology you believe students should be paying attention to right now?

Paul Saunders: Reinforcement Learning (RL) is slowly becoming more applicable to industry. For example, with Martha White’s startup RL Core, it’s getting to the point where RL jobs are feasible. It’s still something you need to be very specialized in, but just knowing about RL isn’t enough. It’s something to keep an eye on and get more familiar with if it interests you.

Okay, that’s fantastic. Is there anything else you feel students should keep an eye out for beyond RL?

Paul Saunders: Not really, nothing niche. Just stay aware of how things are going generally.

Paul Saunders: I think a lot of AI is just hype. There’s a seeming need to put AI in a lot of stuff that isn’t necessarily needed, or to advertise AI in things that just don’t need it. Like, there are toothbrushes marketed as having AI, and I don’t think that’s necessary. I wouldn’t say any specific development tool is overhyped; it’s more a matter of figuring out what works for you. But certainly, some products and the general trend of putting AI in as much stuff as possible, I think, are a little overrated.

You know, especially with agentic AI that’s recently come up, I’ve noticed a lot of people are fearing for their occupations, their jobs. How do you feel about that?

Paul Saunders: I mean, it depends on your role, right? The computer definitely replaced some jobs, but it also made people more productive and opened new possibilities. So, maybe your job will be replaced, but I don’t think your position as a skilled individual will be replaced. You might have to learn new things, but hopefully, you were going to learn new things anyway over the course of your career. So, I don’t think it’s necessarily a bad thing.

What cross-disciplinary skills have helped you most in your career?

Paul Saunders: Yeah, the biggest thing has to be the soft skills—communication and teamwork abilities. I’ve been talking to a few recruiters recently, and they say the number one thing they see lacking in new graduates is that while they have good technical abilities, their soft skills are lacking. Because their soft skills are lacking, they don’t work very well in teams or have an inability to communicate their abilities, and so they don’t necessarily get a job. Having those soft skills, like learning how to work in a team, learning how to work in a software development team via internships or open source contributions, that’s super essential, and it’s also not really taught in typical coursework.

What would you say something that undergrads can do by the time they graduate to actually develop a bit of their soft skills?

Paul Saunders: One would be general projects, but not just like, “Oh, I built a small app.” Open source is probably the best thing to do because it’s an existing, large codebase in which you need to work with others to make changes. A project is cool, but if you can show that you can work in a team and you have a track record of working well in teams, that’s way more beneficial than just walking through a project you did. So, open source is definitely beneficial, and executive work—anything you’re doing where you’re managing something—is super useful.

What’s the most underrated experience or project you did during your undergraduate that helped you farthest in your career?

Paul Saunders: The projects that are somewhat underrated aren’t necessarily projects that themselves were like, “Oh, this is actually a really cool project.” Instead, they were projects that gave me an opportunity to meet other people. For example, I did a few research positions with professors, and those projects weren’t themselves very interesting, but they fostered relationships where those professors knew other professors and would say, “Hey, I know this undergrad or this new graduate student. They’re somewhat interested in what you’re doing; maybe you should talk to them.” Without that experience, I wouldn’t have had those connections. So that’s where it kind of comes down to communication and teamwork.

If you had 60 seconds to give one piece of advice to an aspiring AI student, what would you say?

Paul Saunders: AI is a massive field, so being an AI student is like being an engineer. The field is changing very quickly, but you can’t just say, “I want to learn AI.” That’s kind of like saying, “I want to learn math.” There are so many different subdomains to dip your toe in. And especially because, as I said, the field is changing so rapidly that you need to be aware of what’s going on. So, join reading groups, network with other people, but more than anything, just be aware that you’re never going to really feel like you know everything. And more importantly, you need to constantly be learning.

Do you have a favorite AI resource, tool, podcast, or YouTuber that you normally consume?

Paul Saunders: Uh, I don’t really think I have one. I don’t really watch much AI YouTube or anything like that. Computerphile is cool. In terms of resources, I go to research labs, I talk to people, I find out what they like, and then from them, I find out cool papers they like. I wish there was a really good centralized place to find new papers or anything like that, but there doesn’t seem to be. GitHub Copilot is really good, especially because as a student in the GitHub Developer Pack, you get free Copilot access, so definitely take advantage of that. But yeah, other than that, I don’t think there’s necessarily a big resource that I’m using. I just kind of keep my ear to the ground.