Dr. Theodorea Regina Berry, Vice Provost and Dean of the College of Undergraduate Studies, is joined by Interdisciplinary Studies major Stefan Lunis for a discussion about the future of A.I. and how his time at UCF led to him doing research at M.I.T.
Dr. Theodorea Regina Berry:
Welcome to Academically Speaking. This podcast is designed to provide our listeners with an opportunity to engage with subjects and topics related to student academic success. How we think and what we do is important to how we become citizens of this country and of the world.
Hello and welcome to Academically Speaking. This is Dr. Theodorea Regina Berry, Vice Provost and Dean of the College of Undergraduate Studies. And today our topic is Unleashing Potential: Interdisciplinary Approach to Learning. My guest with me today is Stefan Lunis, who is an Interdisciplinary Studies major, with minors in Cognitive Science and Philosophy. He recently completed a research program with MIT. Welcome, Stefan.
Stefan Lunis:
Oh, thank you for having me.
Dr. Theodorea Regina Berry:
Thank you for joining us today. MIT, that’s impressive.
Stefan Lunis:
Yeah, it’s still taking a little bit for me to let it all sink in, in all honesty. Yeah.
Dr. Theodorea Regina Berry:
So we’ll talk a little bit more about that. But first, talk to me about why you chose Interdisciplinary Studies as a major.
Stefan Lunis:
Well, I’ve been trying to… I’ve been on a pretty long journey to figure out what education path would probably be most beneficial for me. I think I started out as a biomedical sciences major when I first transferred from Lake-Sumter, and then I transferred over to computer science major. But I finally settled in Interdisciplinary Studies with a minor in cognitive science and philosophy, because it just synthesized the ideals that I wanted to touch on and focus on the best so far, as far as degrees offered.
Dr. Theodorea Regina Berry:
So when you think about the things that you are passionate about in relationship to your major, give me some keywords that you would use about what your passions are?
Stefan Lunis:
The big hot button issue, AI, artificial intelligence is something I’m hugely passionate about, as evidenced by my shirt, and repping the AIUCF Club here, and then just a lot of different activities. Ethics, I think play a really big background in the way I want to approach the world. Coming from immigrant parents, from Haiti and Jamaica, respectively, I feel like a lot of the way that I see my own progress will always be in proportion to the progress of others around me. And so when I’m in these new spaces like AI and MIT, I always want to think about the ethical ramifications of the technology I’m developing, and also the people who are going to just be affected by it in general.
Dr. Theodorea Regina Berry:
That’s a good response, and thank you for sharing that. So as you think about the things that you’re passionate about, what kinds of courses do you think have made the greatest impact on your journey, on your academic journey?
Stefan Lunis:
This one’s maybe not as loaded as I would think, but in all honesty, my philosophy courses. My philosophy courses, I was a little hesitant to take because I had always been told growing up that a philosophy degree may be fun, but it won’t get you a job that pays very well. But when I took the courses, there was this electric, a resonance between me and the material where I realized, “Oh, it doesn’t matter if this paid. I love it, and you don’t have to give me a grade for this. I’ll read it regardless.”
And so that understanding, and almost excitement I get from philosophy, and the fact that I can see clearly that there are different people who have been thinking the same way I’ve been thinking, or even analogous to the way I’ve been thinking for hundreds and thousands of years, and that some of these issues that they’ve been grappling with we’re still grappling with today. It kind of gives me some optimism about the current world we live in, that even if we don’t have solutions, there are techniques and methodologies out there that can help us get towards a better world that we’re hoping for.
Dr. Theodorea Regina Berry:
Well, I certainly hope that philosophers and people who study philosophy find their way to spaces. I’m a curriculum philosopher by training, and so I read the works of people like John Dewey, and John Goodlab, William Pinar and the like, and they’re asking questions about knowledge acquisition, knowledge construction, knowledge production, and the value of knowledge in various spaces, and how the context of knowledge changes with different groups of people. And so I find that absolutely fascinating, but I’m that geeky nerd that asks all those questions, and I’m proud to be a geeky nerd, thank you very much.
Stefan Lunis:
Likewise.
Dr. Theodorea Regina Berry:
So what kinds of activities are you participating in outside of the classroom?
Stefan Lunis:
Ooh, this is going to be a pretty long one. So first and foremost, I am the former president of the AIUCF Club, current director of discussions. We actually have a meeting today, at Wednesday, where we’re going over another AI paper, and I’ve been involved with that group for probably about the past three to four years. In addition, there are a few different local activities I try to get involved with. Before I was doing the fun things I do at MIT now, I was working as a bar back and a bartender, a server and a caterer in the service industry.
I like to think that I still have a lot of friends in that community, and I carry a lot of those sensibilities and a hard work ethic that came from there, and it transfers over into what I do now. But explicitly, one of the things I was trying to work on, or I like to work on on the side, is a small… Bespoke AI may not be the right word for it, but small, focused interventions into small, local owned businesses, and using AI to either help them on the backend or help them in a more creative sense.
So currently there’s a bar in Mount Dora called Handlebar that I used to work at, that I’m trying to develop a AI cocktail program for them that takes the base concepts behind most cocktails from, say, a book released in about 2008 called Cocktail Codex, that distills any cocktail you have into six different families and formulas. And then from there you can use those archetypes and see what ingredients you have in your current space, and then see what actions are necessary, and then use that representation symbolically to generate using the current AI we have. Specifically the one I play around with is ChatGPT, because it’s really popular and really robust, and using that to get decent to slightly above average recipes. And then with those recipes, we can then taste test them and see which ones are best, and then use that for a weekly menu.
And then hopefully, from that, we can show that even small interjections with AI into local businesses might be beneficial. And that runs directly parallel to the research I was doing up at MIT, where we were taking specifically representations of really rich domains, a la cooking, and taking them and boiling them down into their base elements through a really interesting lens. If you’ve ever played a game like Cooking Mama, or Overcooked explicitly is the domain we used. We use PDDL, the planning domain and definition language, to represent these actions in a way that a robot conventionally would understand. But it’s not about necessarily putting this in a robot, but about the fact that PDDL has this really structured format that we can then use. And then what we ended up doing was taking these Overcooked styled recipes, translating them into PDDL, and then giving them to GPT3, which then gave it back to us in PDDL.
And then testing humans to see, one, how well these generated recipes compared to human made recipes, or even a random baseline of just… My favorite example of a random baseline is it’s technically possible and feasible, but it’s not necessarily something you would want to do. We rated it by quality, feasibility, and oh, wow, I’m forgetting the last metric. But a good example would be if you take onions, lettuce, tomatoes, you mince them into bits and you put them in a pot, and then you take that pot and you bake it in the oven for five hours, and then you serve it on a cutting board. That’s possible, but not very likely. And comparing that random baseline with, say, a pizza with anchovies, versus say, a pizza with anchovies and pineapples. It might be a little easier to tell which one’s the human and which one’s the AI.
But if you do that enough times, you see a bit of a trend. And we then also tested to see how those recipes, when shown bit by bit, like someone’s in the kitchen cooking something, which recipe lines up best with what they believe was being generated. And through that research, we hopefully are tapping into this common structure that these general large language models are training on, because they’re training on so much information, and hopefully we can pull from these structures these base, human biases towards certain recipes, or ingredients, and we can learn more about the way we learn. And I want to take that high end perspective of learning the way that we learn from these simple and sometimes complex tools. And then taking these applications and putting them directly in small business hands. And then also taking this information I learned and bringing it back to my community at UCF through the AIUCF club. And I run the gamut, holistically, between those three points constantly.
Dr. Theodorea Regina Berry:
But there’s always going to be some challenges in relationship to an AI developing something from a recipe, versus the way an individual will make something. So for instance, in my family, we make conch salad, and one of the key pieces that I’ve learned about that is that if you don’t have fresh conch, it’s never going to be the same, right?
Stefan Lunis:
Absolutely.
Dr. Theodorea Regina Berry:
And so I can make it in The Bahamas easily, because there’s fresh conch, but if I ship that conch from The Bahamas to Miami, and then get it to Orlando, and then make it, the same ingredients, the same amount of the ingredients, the same process for making it, but it’s not going to taste the same.
Stefan Lunis:
Exactly.
Dr. Theodorea Regina Berry:
As opposed to, in baking, because baking is more of an exact science, it’s more chemically based in measurements, I could see where an AI could actually take the same set of measurements with the same ingredients, and create something that’s precisely the same as if Martha Stewart had baked that cake, right?
Stefan Lunis:
That’s a beautiful perspective, and I agree wholesale. I think when it comes to these representations, when you are doing something like baking, like you’re saying, it’s much more calculation focused and it’s much more precise. And even though there’s room for error, you might get something more Martha Stewart-esque. But I think that touches in on why, when we bring them into real human situations outside of the research lab, why we have to tweak these designs a little bit. Because when we take these concepts that are developed in these really beautiful and smart places, full of smart people, and then release them into the wild without any consideration for how fresh the conch might be, to take this analogy, then we not only shoot the chefs in the foot by saying, “Make this,” and it’s not going to be the best product. We’re then assuming in this, to run with this analogy, that everyone who’s providing conch for all these AI recipes understands… How do I phrase this? Understands really what they’re dealing with.
They’re not dealing with someone who understands the semantic understanding behind the recipe, but instead they understand the novelty of each individual ingredient, and how they may pair. Which is why I really love the cocktail analogy, because with cocktails, one, you can taste it really quickly and figure out if it’s good or bad, but with cooking, there’s so many steps that can go wrong. As opposed to cocktails where, if you are, say, making three different recipes, it may take 10 to 15 minutes, versus say a few hours for a few different recipes in cooking. But then you can taste test very quickly, or even in small samples, what works versus what doesn’t work. Because as long as you have the human in the loop, that improves the process entirely, which is actually how we got to the current ChatGPT landscape that we’re in right now that people are so excited about. Because of our-
Dr. Theodorea Regina Berry:
Some people are excited about.
Stefan Lunis:
Oh, some. Yes.
Dr. Theodorea Regina Berry:
Right. But if you talk to some of my faculty colleagues, they’re not so excited about it. But this is where this notion of interdisciplinarity varies.
Stefan Lunis:
Absolutely.
Dr. Theodorea Regina Berry:
So depending on your discipline, depending on your perspective, depending on your knowledge base in relationship to that discipline, you will have a particular perspective about the advantages and the assets connected to ChatGPT.
Stefan Lunis:
Absolutely, absolutely.
Dr. Theodorea Regina Berry:
Versus the disadvantages. And so I’m dealing with both of those things as Dean, because part of my identity as an academic dean with academic programs, is about what kinds of things would be beneficial to students in relationship to this learning tool? Because I’m seeing it as a learning tool, simply, right?
Stefan Lunis:
Absolutely.
Dr. Theodorea Regina Berry:
But as an administrative dean who’s dealing with academic policy, the question becomes what kinds of challenges are we faced with around academic integrity in relationship to students’ use of this platform? So now I’m seeing it as a platform rather than a learning tool, in relationship to student activity and engagement. And like the cocktail, it’s about the difference between whether or not you’re using a potato based vodka versus a rice based vodka versus a chemically generated vodka, and whether or not it’s going to cause that cocktail to taste differently.
Stefan Lunis:
Incredible subtlety. And that also opens the door to the interdisciplinary nature of the research inherently, and how sometimes outside of the areas it’s developed that can be lost. The labs I work in are Josh Tenenbaum’s Computational Cognitive Science Lab, as well as Vikash K. Mansinghka’s Probabilistic Computing Lab. And other than both of those things being a mouthful, most people say instead of computational cognitive science, they call it CoCoSci, because it’s cuter and it’s a little sweeter, but from its very invent, or from it’s very origin and its ethos is so interdisciplinary, that it felt like I was moving in a perfect trajectory to where I was supposed to develop when I got a chance to research up there.
But taking these tools outside of these really vast, like philosophy, anthropology, linguistics, neuroscience, and even robotic approaches, all within the same lab sitting next to each other, talking to each other, and throwing it into the wild, and assuming that everyone will also have those contexts… This is a little naive. Which is why we need to not only make it more, I guess, more accessible, we need to understand its ability to be used as almost like an equalizing platform, as you’re saying. My best comparison is when Wikipedia came out, when I was… This might be dating me a bit. When it came out in middle school-
Dr. Theodorea Regina Berry:
And everybody was saying, you can’t use it because it’s not valid knowledge, right?
Stefan Lunis:
Exactly. And while ChatGPT, I would say for anyone using it, either within the academic code or outside of it, the accuracy varies wildly, as someone who’s used it in research and outside of research myself.
Dr. Theodorea Regina Berry:
But what’s going to get interesting is when they merge with Google Scholar, right?
Stefan Lunis:
Of course.
Dr. Theodorea Regina Berry:
Because then you have something that’s already been established as valid knowledge, in a platform that has still some controversy attached to it.
Stefan Lunis:
And the way that we get to that point is by exposing people to the tools we have now, ChatGPT, and understanding that it’s a work in progress, but then it doesn’t just magically… That’s a great point. It doesn’t magically get from ChatGPT to ChatGPT with Google Scholar. People need to develop these tools and integrate, AI’s not at the point where it’s just reaching and grabbing and pulling things in yet. We need to pull it to that place. But the only way we get to that place is by making it more equitable. So everyone not only has access to it beyond paywalls and things like that, but that everyone has exposure and familiarity with these tools. I want to see a world where people are using AI the same way that I saw a two year old the other day, watching and using an iPhone, and I’m like, “They’re doing better than my 100 year old grandpa on this iPhone.” And it’s not because that they’re just smarter, it’s because specific-
Dr. Theodorea Regina Berry:
It’s intuitive.
Stefan Lunis:
It’s intuitive, and it moves along with their worldview. And we, even specifically in the lab that I’m in, we discuss some fundamental, or these concepts that we’re dialoguing about, like intuitive physics and intuitive psychology. That from a very young age, we have some fundamentals about the world that we interact with that may be ingrained, versus even just say learned. And I feel like if we come in at the right window, we can, for children and even people in college age right here and right now in this AI revolution a bit, we can learn to put these tools in the hands of people who won’t only just familiarize themselves and get better at using it, because the more the merrier. But specifically whenever there’s a new technology, and this is a bit of my favorite soapbox, I guess. I’ll stand on it for a moment.
Whenever there’s a new technology, say railroads, electricity, automobiles, usually what they say in Cambridge, URM, underrepresented minorities tend to be the people who get hit the most. Or just disenfranchised folk from the beginning. If you look at the railroads, who’s building it? Chinese labor workers. If you’re looking at the automobile with Henry Ford, who’s building it? Children in assemblies and factories. And the same thing happens with AI, where you see people who are developing these AI, and what do we call it, the AI white guy problem, where it’s like, because of the homogeny of people within these labs and rooms… Usually. Luckily, my two have been a pretty good exception to that rule. You end up with AI that’s not inherently evil or mean or malicious, but looks at Black people and thinks they’re gorillas because they literally haven’t been training enough Black faces because they thought that the celebrity database was sufficient enough. Because they’re like, “Well, I see all these celebrity faces and they’re only a few, but that’s just the way the world is.” It’s like, well, the world’s pretty large.
Dr. Theodorea Regina Berry:
Right. So we’ve spent a lot of time talking about this.
Stefan Lunis:
Sorry.
Dr. Theodorea Regina Berry:
No, but it’s been a great conversation. But I want to learn a little bit more about you.
Stefan Lunis:
Absolutely.
Dr. Theodorea Regina Berry:
As a student and a person. And first of all, getting back to MIT, how did you get to MIT, right? What opportunity did UCF present to you that afforded you the chance to study at MIT?
Stefan Lunis:
This might sound a bit like a wild story, but I’ll start at the beginning, which was, I think I may have just transferred from biomedical sciences where I was in a class at UCF. I forget what the class was, but they exposed you to all the different kinds of doctors and different kinds of medical fields you could focus and hone in on. And at the last lecture, it was the MD who came and talked about how he had an undergrad degree in, I think Eastern Philosophy, then had to get a master’s degree, and then go to med school, and went through a lot of ups and downs. But at 40, he established himself and settled. And I remember him talking about the things he loves doing, and how he’s using AI and anomaly detection in lungs and things like that.
And I was really excited about what he said, and he seemed really happy. But then I walked away thinking, “Wow, I love this, but I don’t want to wait till 40 to do the things that I want to do. I think I’ll be a pretty decent doctor,” but I mean, when I was a kid, well, from three years onward, I’m like, “I’m going to be a doctor and a pastor.” Those two things. And then obviously they deviated a bit, but I then immediately transferred to computer science, and I didn’t really know what to do. No one in my family had done computer science at that point. And so I just went to my local AI club, thinking, “Well, I don’t know how this works, but I got to be in that room where it happens. If this is moving around me, I want to be a part of the process.”
And from attending those meetings and familiarizing myself, I learned that every once in a while, a few people from MIT come down and scout in these areas, and they like to look even in AIUCF and a few different clubs. And then if you meet them and do well with them, and have a pretty decent portfolio, they’d be willing to give you an opportunity to do different programs. Harvard does this as well. But the one that I went to for MIT was the Massachusetts Summer Research Program, specifically within bio, because computational cognitive science falls within cognitive science, which falls within the broader category of BCS.
Dr. Theodorea Regina Berry:
Sure.
Stefan Lunis:
The biological sciences. And that opportunity came by working service jobs for a long time, until eventually through my connections to the AI club, I was able to do AI and ML at Photovoltaic Solar Energy through the Florida Solar and Energy Center, which is based out of this… I guess what, the Sun Coast? But specifically out of UCF as well. And Dr. Hubert Smith, an incredible mentor helped develop my skills to be a good researcher so that when they came down the next year, I got to interview and talk.
And they’re like, you have the credentials, you have the letters of rec, let’s give you a shot. And then once I got up there, I learned so much that they just kept letting me come back. They’re like, “Oh, you want to come back for the fall? Sure. You want to stay a little bit longer in January? That’s okay.” And they’ve just been so conducive to me that even though I didn’t feel like I would’ve ever had a chance to do the GPA requirement, once I developed my base down here, it gave me a foundation to build my confidence and my skillset enough that they acknowledged what I had done. Hopefully that answers the question well.
Dr. Theodorea Regina Berry:
Oh, absolutely. Absolutely. So outside of the classroom, you mentioned you’re in the AIUCF Club and you’re doing a few other things. So what advice would you give to your fellow students about why you participate in student organizations, why you do research, and why you took advantage of the opportunity to study at UCG? What advice would you give to your fellow students?
Stefan Lunis:
Ooh, one, don’t let your GPA scare you away from doing something you think might be really beneficial. A lot of the times they’ll say on paper that they have a hard requirement, but if you meet with someone and talk with someone in dialogue, usually they might either be a little more lenient or understand your circumstance. Once I explained my own individual situation, paying my way through school, it became a lot more clear that it wasn’t that I was some bad student, but that there were adversities you have to face in life. And they definitely can be empathetic towards that.
In addition, just seeing options and pursuing them, I think is incredible. It may feel as though the goal you’re setting now to accomplish is unreachable or too farfetched, but as long as you’re putting yourself in the situation, in situations and scenarios where other people are achieving these goals, you’re most likely going to acquire not only the tools and the techniques, but the mentality that allows you, and as weird as this may sound, the language that allows you to move in and out of these spaces, and allow you to not only be accepted, but to thrive in these situations.
Dr. Theodorea Regina Berry:
Absolutely. Okay. So now this is the fun part of our conversation, which is what I call the speed round.
Stefan Lunis:
Go for it.
Dr. Theodorea Regina Berry:
I’m just going to ask you a set of questions. You answer me with the things that come to the top of your head, okay?
Stefan Lunis:
All right.
Dr. Theodorea Regina Berry:
All right. Favorite color?
Stefan Lunis:
Red.
Dr. Theodorea Regina Berry:
Favorite song?
Stefan Lunis:
Ooh. Ooh, that’s hard. I want maybe Rubberband Man by The Spinners, because I loved it as a kid.
Dr. Theodorea Regina Berry:
Oh my gosh.
Stefan Lunis:
Yeah.
Dr. Theodorea Regina Berry:
You’re an old soul.
Stefan Lunis:
Yes. Very much so.
Dr. Theodorea Regina Berry:
Okay. All right. Favorite movie?
Stefan Lunis:
Ooh, I probably… Oh, wow. Either Everything Everywhere All at Once, which I saw recently, or perhaps Fight Club. Not necessarily for the current connotations it has, but because I was 12, and it blew my mind with the twist, and I’m like, “Wait, what?” And that stuck with me to this day.
Dr. Theodorea Regina Berry:
Favorite TV show?
Stefan Lunis:
Ooh, I love comedy. Oh, so maybe Community or even Abbott Elementary. But The Good Place is probably my favorite TV show, because if you can’t see me through the audio, I do look a little bit like Chidi from The Good Place. And I have a lot of the same anxious and philosophical quandaries, so I get called Chidi a lot.
Dr. Theodorea Regina Berry:
Favorite actress?
Stefan Lunis:
Ooh, favorite actress. Oh, wow. I always used to say Meryl Streep just because of her range and the lack of Oscars she deserves. So I’ll stick with that.
Dr. Theodorea Regina Berry:
Okay. Absolutely. Favorite thing to do in your free time?
Stefan Lunis:
Ooh, currently play bass guitar. That, or occasionally play video games.
Dr. Theodorea Regina Berry:
Okay. What’s your favorite video game?
Stefan Lunis:
Oh, wow. Anything a really good RPG or anything in this genre called rogue-likes, where you’re playing a game, and then you die a lot and then you get brought back, like Hades or Slay the Spire. I think there’s a… Or Elden Ring. There’s a really visceral response I have to playing a game, losing, and it saying, “Don’t worry, losing’s part of the process,” and then continuing along. And it gives me this satisfaction that I also get from my education career where I’m like, “All right, there will be losses, but you get stronger for it at the end of the day.”
Dr. Theodorea Regina Berry:
Yes, yes. All right. And so I’m going to end this interview with a question that I’m taking from a TV show that I’ve been watching recently called If We’re Being Honest with Laverne Cox.
Stefan Lunis:
Oh, I love Laverne Cox.
Dr. Theodorea Regina Berry:
She’s awesome. So at the end of her interview, she typically asked her guest if there was something that she should have asked that she didn’t ask. So I’m going to ask you, is there something that I didn’t ask that I should have asked?
Stefan Lunis:
Ooh, that’s a heavy one. I don’t think there’s anything I can think of that you should have asked that… Or that you could have asked, that you should have asked. If I have to speak of anything more that I would’ve wanted to be prompted on, oh, my family means the world to me, and I literally wouldn’t be here without them, or my mentor Chenjie Shen up in MIT, or just any of the people who… I guess the question would be, who got you to this point? Because I can’t think about where I am right now without thinking about every single person who got me to this point. I literally just celebrated my… This is definitely dating me. My 30th birthday two days ago.
Dr. Theodorea Regina Berry:
Oh, okay. Happy birthday.
Stefan Lunis:
Oh, thank you. Thank you. And all I could think about was how many various different groups of people, ups and downs, made me who I am. And I’m so grateful for the place I’m at right now.
Dr. Theodorea Regina Berry:
And that is a great way to end our conversation. Thank you for joining us today on Academically Speaking.
Stefan Lunis:
Thank you for letting me be here.
Dr. Theodorea Regina Berry:
And thanks to our audience for joining us. This is Dr. Theodorea Regina Berry with Academically Speaking, and have a great day.
Podcast: Download