Subscribe
Share
Share
Embed
Brains, Black Holes, and Beyond (B Cubed) is a collaborative project between The Daily Princetonian and Princeton Insights. The show releases 3 episodes monthly: one longer episode as part of the Insights partnership, and two shorter episodes independently created by the 'Prince.' This show is produced by Senna Aldoubosh '25 under the 147th Board of the 'Prince.' Insights producers are Crystal Lee, Addie Minerva, and Thiago Tarraf Varella. This show is a reimagined version of the show formerly produced as Princeton Insights: The Highlights under the 145th Board of the 'Prince.'
Please direct pitches and questions to podcast@dailyprincetonian.com, and any corrections to corrections@dailyprincetonian.com.
Senna Aldoubosh 0:14
Hi everyone! Welcome to Brains, Black Holes and Beyond: a collaboration podcast between the Princeton Insights newsletter and The Daily Princetonian. From the 'Prince,' my name is Senna.
Lina Kim 0:23
And my name is Lina.
Senna Aldoubosh 0:24
Today's guest on the show is Dr. Pedro Parades, a lecturer in the computer science department, Dr. Parades, got his bachelor's and master's at the University of Porto, and his PhD just recently in 2022 at Carnegie Mellon University. He now does research in primarily theoretical computer science, runs groups like the Princeton Competitive Programming Club, and sits as the head of the scientific committee of the Portuguese National Olympiad. Welcome to the show!
Dr. Parades 0:49
Thank you very much! It's a pleasure to be here.
Lina Kim 0:52
So I guess we'll just get started off: what initially got you interested in computer science, more specifically, theoretical computer science?
Dr. Parades 0:59
Yeah. So, if I recall correctly, when I was in middle school, my math teacher brought a former student of hers to the school to kind of teach people about programming. And back then I was interested in computers already. I was very interested in things like 3d graphics and 3d modeling — so this was very natural for me. And I got really interested in building things with programming. I remember I built like, you know small robots. But then I got interested in kind of the mathematical aspect of it through a kind of puzzle, or logical puzzles, which are things that we kind of... almost everyone has seen puzzles, and like, they're kind of frustrating. Sometimes they're kind of fun. And I thought, I think it was something that I still think a little bit about my work is kind of like solving these like mathematical puzzles. And in particular, I'll leave one puzzle for the listeners that I remember being very interested about when I was in high school later, and always kind of when I was kind of doing a little bit more math. And I quit — and I was not able to sell this back then. But it's a very interesting puzzle, which is usually called the "100 prisoner puzzle." It's kind of a popular one, maybe some of you have heard it before. I'll tell a happier version of it because the original puzzle is a little dark. But let's say there's 100 people, and you're playing this game where there's 100 boxes, and each box says someone's name. And it's 100 people in one room and the 100 boxes that are in a different room. And the 100 people have to come up with a strategy, so that they'll walk to the room where all the boxes are, one by one, independently. And they have to find the box that contains their name. And they have only 50 trials. So, there's 100 people, 100 boxes, and each person has 50 tries to find their name. And they have to come up with a strategy, they can come up with that strategy beforehand. But then once the game kind of starts, they can't talk to each other. So each one of them will go into room, they can look at 50 boxes, and then close them again. And they have to find their names. And if they all of them, every single one of them finds the box that contains their name, they win. So, this is the goal. And the goal is to try to come up with a strategy that won't work all the time, but with work with like a pretty high probability. And for example, if you just do the obvious thing, where each person will pick like 50 random boxes and look at the 50 random boxes, each person will have about a 50-50 chance of finding their box. But then for 100 people to find each one of their boxes, the probability of each one tossing a coin and landing on heads. And you have to have 100 people do that, which is very, very unlikely. So we have to come up with a strategy to do this, that has a much higher probability than this strategy. For example,
Senna Aldoubosh 3:26
I have a follow up question.
Dr. Parades 3:27
Okay.
Senna Aldoubosh 3:29
Does each person — like when they go into the room — do they have to do all 50 of their turns when they go in? Or is it like, can one person go in and do one and then leave, and then come back later?
Dr. Parades 3:40
Yeah. So the first one. One person comes in looks at 50 boxes, and then that's it, then they leave and they can't talk to anyone else. So it's like, you can coordinate initially, but then you can't talk to anyone else anymore. Someone goes in, looks at their boxes, leaves, and then repeat.
Senna Aldoubosh 3:54
And there's no way to find out whether or not that person actually found their name.
Dr. Parades 3:57
Well, if someone doesn't find their name, then everyone loses the game. So, the game could basically stop there. The goal is to get every single person to find your box or the box that contains their name.
Senna Aldoubosh 4:06
I see. That's a very fun puzzle.
Lina Kim 4:09
I'm like thinking about it... wait.
Dr. Parades 4:11
I lost a lot of sleep because of this puzzle. It's a very fun one. And if you look up online "100 prisoners, 100 boxes" you'll find that puzzle and solution.
Lina Kim 4:21
Very interesting. Yeah, hearing that definition. Now, I feel like my view of theoretical computer science has definitely changed. So I guess that being said, what is something that is misunderstood in your field? Or any kind of misconception that the general public you think may have?
Dr. Parades 4:36
Yeah, that's a very good question. There's a lot of things I could talk about. I think, theoretical computer science, because it's the theory of computing and kind of touches on almost every field of computer science. One good example that I like to think about that is related to some of my work is the in the field of quantum computing, which is something that a lot of people have probably been hearing about recently. It's a very hot topic in computer science, in physics and math. And sometimes in the media, it's kind of thought of as a big solution for all problems. Like if we have a quantum computer, we can kind of... computer predicts a lot of amazing things. And even though this is true, like everything in computer science, there's limitations. It's not a magical power that will allow us to do anything. It will allow us do a very specific set of things. And even those things, we're not 100% sure if they're not doable by a normal computer. When it shows up in the news, they're always kind of overblown. And quantum computing is an example of that. But I could go over many, many examples that have this property.
Senna Aldoubosh 5:35
Gotcha. I'm talking a bit about like misconceptions, I guess, where do you see the future of society headed as we see computer science become more prominent in all aspects of life? I know it is already, but like, kind of extending it to the future a little bit more?
Dr. Parades 5:47
Yeah. It's a very good question. And I think it's the right time to ask this question, because everyone has probably seeing the news and seen how there's these new tools like Chachi, PT, and all these large, large language models that are being released by several companies, it's very hard to predict where things are going to go in the next 5, 10, or 20 years. I think we find ourselves at a place where looking at — if we think about kind of where we are as a graph, or as a little plot, it's going up very fast. And it can either explode and we can get to a point where in a few years, artificial intelligence can kind of do almost everything. It can do things that we never thought it'd be able to do. Or it could stagnate — kind of stay where you are. And I think the next few years are going to be very critical to see what is the power of these modern tools of machine learning and computer science. And to make this a little more precise, for those that haven't played with things like ChatGPT, or like these large language models. These are models that collect a lot of data over the Internet, for example. They have a lot of textual data that they collect. It's a huge data set that is collected on basically almost anything that a human could have written that is available publicly, for example, online. And then by collecting all of this data, these algorithms have a huge predictive power to guess what the answer to that prompt or like, the next words would be given in all the data kind of extracting patterns from your data, and try to complete it according to what the prompt was. And one of the big, kind of, there's a lot of computer science behind these things. And there's a lot of reasons why this is something we are able to do now and not, for example, 10 years ago, related to also some hardware constraints, for example. Now, our computers are much more powerful than it used to be a few years ago. But there's also some algorithmic or some fundamental ideas that came into these models that kind of use context in a way they didn't use before. Because I can like we can play this game where if I say a certain word, you can kind of guess what the next word is. If I say "Princeton," you probably can guess that the next word is going to be "University." But if I say a full sentence, it's harder to predict what the — after a couple of words — it's hard to predict what the next word would be. You're kind of just saying random words that connect to the previous word. But that's not how language works, right? We have context, if we're telling a story, there's a few characters that show up. There's a lot of memory involved with it. And these models are kind of able to mimic this memory aspect. And so they're very powerful, and they are able to generate text that really sounds very human-like, in a way that we haven't been— wasn't possible before, at least we haven't seen before. And so these models that have a lot of power to basically mimic human speech, and potentially with more work and with more developments, they might be able to do things like, for example, math. And they already are very good at for example, programming, they're surprisingly good at programming short-ish pieces of code. So they can write kind of like small programs to do certain things, which is something that was before kind of thought of like, "oh, we only humans can do this." And so it's very exciting and scary at the same time to see what these things can do in the future because as they get more powerful, and more powerful, they might be able to do certain things that we never expect these models to be able to work machine learning in general to be able to.
Lina Kim 9:16
Yeah, I mean, all of that sounds very valid. Definitely. You know, with the future, we got to be mindful of those kinds of things when it comes to like ethics and AI and whatnot.
Dr. Parades 9:25
Let me, since you mentioned that, one thing that I think is important to say is that us as a society are not really ready for these tools, even if these tools are not as powerful as we thought. As everything in computer science, there's an impact in society. And usually the researchers and computer scientists are working on these models are more interested in the technical aspects, or computer science behind it. Which makes sense because that's their primary job. But then the implications that these models have in our society. There's a lot of people to think about these, but it's kind of an area that is not as explored as it should be. And there's a lot of potential implications that they weren't expecting. And we see that a lot nowadays with, for example, how these social networks have had an impact and kind of can influence people at targeted advertising. Basically, everything that we do online, nowadays, and we do with our phones or with our computers, has an immense power that we weren't really expecting in the past. When people were building these tools, they were just building tools that they thought were going to be cool for people to use. And the same thing will probably happen, or most likely happen, with these new AIs where they'll have consequences that no one would expect, or maybe some people would expect. But it's sort of like those movies where there's someone like screaming, "oh, we have to save the planet, we have to stop doing this." And the politicians ignore them and society ignores them. And that kind of thing is probably happening right now.
Lina Kim 10:47
Yeah, no, definitely. I guess with all that being said, Is there anything else that you would like to include in the podcast or any general advice you'd give to listeners?
Dr. Parades 10:57
Yeah, my general advice is that for everyone that is now in college, or even after college, and are thinking about their careers and thinking about their lives, I think it's very important to learn just basics of computer science. Even if you're not really going to use that in your professional life. Clearly we're getting to a point where almost every job requires computer science, so it's always nice to know that. I think it's also important to learn, so that you can kind of protect yourself and kind of know what you're doing. If you're going to interact with these tools, you're going to interact with the Internet, you're going to interact with computers. And I think it's very important, and is increasingly more important to be aware of how these things work. And for those that are already kind of doing computer science, for example, people who are majoring in computer science or some adjacent field, I think my main advice and my hope is that as all of you are getting, going to get, or already are in industry or in academia, and like developing kind of the tools for the future. I hope that you also think about the societal implications. Don't forget about the ethics and the human aspect of computing. It's very easy to get lost in the math, because the math is a lot of fun. But I hope that more people think about the implications of their work.
Senna Aldoubosh 12:11
Gotcha. Well, it was really great having you here, Dr. Burns, as we've learned a lot about theoretical computer science. And I feel like before coming into this, theoretical computer science was super intimidating, and I had no idea what was going on. But now that we've kind of talked I feel like it connects a lot more to my life. And I feel like everybody listening's life than we've initially thought. So, thank you so much!
Lina Kim 12:33
Thank you so much!
Dr. Parades 12:33
Thanks for having me!