What one student thinks: Will Krasnow

An interview with the founder of SwiftScore

Swiftscore is a grading application for teachers created by a group of students.

In 1979, this guy riding a horse ran for student body president at the University of Georgia on a platform to disband student government. I arrived some years later and served two terms as president of the University Union. The Union is not as radical as it sounds. The group’s purpose is to spend student fees to bring speakers, bands, and art exhibitions to campus. My favorite was choosing the films for the theater in the student center, the “student union.” 

Since there was no student government president, I got invited to a lot of meetings with high-level administrators where I was supposed to represent students. When a dean or vice president would ask me, “What do students think about this?” I always said, “I don’t know. I can only tell you what one student thinks.” I tell that story as background for this interview to emphasize that this conversation is about what one student thinks. 

Will Krasnow is a junior in the College of Arts and Science at the University of Pennsylvania majoring in Philosophy, Political Science, and Economics (PPE). He founded SwiftScore, which uses generative AI to help teachers grade.

On April 19, we talked for about 45 minutes on Zoom about a wide range of topics. With Will’s permission, I recorded our conversation and used it to create this transcript. I edited it for length and clarity. Will reviewed the transcript and provided edits for a section he thought was unclear. We agree it captures the essentials of our conversation.

---

Why PPE?

Studying politics, philosophy, and economics (PPE) is a bit of a rare thing for someone involved in the tech startup industry. But I'm a big proponent of liberal arts education. Originally, I was super interested in politics and student government. I was the class president here at Penn for my first 2 years.

I realized I wanted to build, to be involved in the private sector, Silicon Valley, LLMs…everything that's going on right now. 

What led you to that realization? That shift in focus? And why edtech?

I'm a big believer in the importance of government, representative government. But I didn’t want to only represent others. I wanted to build. I wanted to have my own voice, a way to make my unique contribution. I think the private sector provides that opportunity. 

Are you familiar with Y Combinator?

Sure. Paul Graham. Sam Altman. Venture capital. Start ups. 

They have an event tomorrow in Boston, and all the group partners and I are going up. We have a bus of Penn students going. I feel like they have done a really great job of turning building a startup into a sort of formula. And one of the things they always say is like the best founders find a problem in their life or a problem in a friend's life, and then just solve it.  And really, my life has been all about education, right? I mean, there are some other things. But the things I think about and care about and experience are related to class and teachers. Edtech was a chance to build something that I know could be useful.

My friends who are TAs [Teaching Assistants] spend a ton of hours grading and would say “Why can’t ChatGPT just do this? And then I looked into it and was like. Well, ChatGPT is not quite perfect, you know. It's not always accurate. But then, could we make it accurate? And I guess that's been the question for SwiftScore.

I've noticed that there is a lot of interest in using generative AI for education. Teachers and students think it can help accomplish their work, but when you ask students about teachers using it to grade or write letters of recommendation, or you ask teachers about students using it to write or complete assignments, there are a lot of concerns.

SwiftScore is right in the middle of this divide. It is made by students for the classroom. But it is a tool for teachers. Some students might feel like it lets teachers abdicate their responsibility to grade. How do you see that? What's your take when a student or a teacher says AI is okay for me to use, but not for you?

I think that's what makes all this stuff so funny. I was talking to a high school teacher. Because we're [SwiftScore] focusing primarily on high school teachers right now. They have a huge workload when it comes to grading. He's like, I love this. It will save me a lot of time.

He said, what if my students are just writing essays using AI, and then I'm using AI to grade those essays, and then…what are we doing? What's the point?

And my take on it is, if that's the case, we probably shouldn't be assigning those things. 

“ChatGPT can’t kill anything worth saving” is how John Warner put it. 

Right. The LLM technology is a paradigm shift in what we're able to produce. We should reevaluate everything…what skills are inherently human? And what are the skills that should be be automated? And how can we actually build AI into the classroom in a way that helps us do the things we are good at? I think SwiftScore, at least for now, works well for grading questions that are for proficiency, for specific knowledge, as opposed to more abstract ideas. We’re not grading art.

I would like it to assist teachers in critical evaluation, but as a co-pilot. I haven't talked to a single teacher who said, “I want to just plug my students’ work into AI, and then, that’s it.” They've looked through the graded questions, and they make changes to the outputs. What they do when they grade is all somewhat subjective, but SwiftScore speeds up the process.

I think that's okay. If it happens on the other end, too. If I'm writing an essay for a class and I talk to ChatGPT and it gives me data I could cite or suggestions on format, I think that's okay. But I had better be writing an essay that's better than whatever ChatGPT could produce.

So tell me more about your educational journey. When you think about the paradigm shift you mentioned and your experiences as a student in arts and sciences, how do you think that has shaped your approach to building technology?

A friend recently asked me a similar question: "Did your classes help you build this?” I had to think about it. I was like, honestly, no, I don't think so.

But I think what I learned is going to help make sure the tech I build does more good than bad. Never in my PPE classes did I learn startup skills: how to do sales, or how to find a co-founder. But they have given me a broader picture of the world and helped me orient the product in a more useful or humanistic way. For example, one of the things that I really value about SwiftScore, in fact it is our whole pitch, is that we help students spend more time with teachers and help teachers spend more time with students. 

There are a lot of edtech tools that are coming out right now that use generative AI to help teachers generate assignments or have students learn directly from the AI. It will ask questions and review answers, and have a back-and-forth exchange where the teacher is not involved. 

What’s the point of that kind of product?  What good comes out of it? I think teachers should be working with students. Students should be learning from teachers. I think that perspective is informed by the liberal arts?

Tell me about a philosophy you've encountered, or a class you took, or a book you were assigned. Something that shaped your view of the world in the way you were just talking about.

A few classes like social psychology, Intro to Psychology, and a behavioral decision science class, which was all about norms and values. I think those gave me a good sense of the human condition, the perils of the human condition.

I'm in a history of education course right now with Professor Jon Zimmerman. There are only six students in the class so twice a week we just meet up and talk.

I think that class has helped me contextualize the different aims of education and how to look at problems from a historical point of view. I don't think I realized this until maybe this year, but to have a good take on the way things should be, you have to know how you got there. What happened historically is the basis for a view into the future. 

Note: I didn’t realize Will was taking history of education, and he didn’t know that I am an educational historian, so the conversation took a turn into details of the class he is taking, the class I teach on the history of higher education in the US, which Jon Zimmerman also teaches. Then we got into philosophy. Will brought up the influence of rationalist thought as defined in the Less Wrong and Slate Star Codex discussion forums, which have been very influential in Silicon Valley and among those who want to work there.  

It sounds to me like you're somewhat skeptical of that rationalist community because it lacks awareness of historical context. They want to start from first principles, not from history or society. Do you see a contrast between where you're coming from and where that rationalist community is coming from?

They're making huge claims, right? These are some of the most powerful people in the world and they are building the technology of the future. And then making normative statements about how the world should be. But it's not like they're philosophers or have a PhD in philosophy. They're sharing their thoughts, and they're very smart people, but they've not been trained in the same ways or adhere to the same processes that great thinkers use, That's not necessarily the worst, but it is a bit of a red flag for me. The second thing I'd say is that thinking like a true historian, we would say there's no such thing as first principles, right? Things change in different contexts, cultures, and time periods. 

I went to a talk organized by Adam Grant [one of Penn’s best-known professors] that featured Kara Swisher. Do you know her stuff?

Yes, Top tech journalists. Has a new book out, Burn Book. A memoir, I think. 

I didn't know of her until the talk. She's awesome. One of the things she was saying is that these tech people's circles get smaller and smaller, and they interact with fewer people. They like to think of themselves as the guy fighting the system. And she's like, No, like you are the system. And if they're making claims from first principles and they’re in this insular community, it's like, where did you get your principles? Where are you coming from? I think that's dangerous.

So you're going to Boston this weekend for a Y-Combinator event. You have founded a technology company. You are looking to join this community. You describe yourself as a founder. Square that circle for me. Let's say you succeed. You get funding. SwiftScore is a hit. How do you take the next steps with this critical framework in mind?

That's a good question. At the end of the day, I am a capitalist. I believe that good often comes out of these tech advances. Look. I took a Marxism class last semester. I took it because it came up in my modern political thought class. 

You don't really hear about it in America. When I went to Mexico City, it was different. They had all these Marxist authors in the bookstores. So, I thought there’s something here, so I took a Marxist class.

It's intense stuff. I think Marx actually had a really good understanding of how capitalism worked. And even he would be a proponent of the idea that capitalism does some good. It produces many things, and then he also thought it would produce its gravediggers. Which hasn't quite happened yet.

One of the things I took away from the course is that capitalism has a velocity of production that no other political-economic system can match. Because of capitalism, global poverty rates are going down, we’re feeding more people, and our comfortable middle-class lifestyles in the developed world can exist. I’ll be ready to discuss a communist revolution when these needs can be globally met without capitalism. 

In other words, I think a lot of good comes from capital valorization. At the end of the day, the people at Y Combinator or the people in Silicon Valley–these people who are driving the innovation part of this machine forward–play a very important role. In SwiftScore’s case, we’re helping address a teacher shortage crisis that is leaving underpaid teachers deeply overwhelmed and students lacking in-person learning.

To solve these issues, the best way for me to do it is to be involved with Y combinator and those tech communities.

However, unlike many of my tech peers who are writing for Less Wrong, I will consult academic experts as I help create this future.

Even so, I understand a lot of bad comes from capitalism. Marx did a great job showing capital’s tendency towards extreme negative externalities. 

That’s where I turn to social entrepreneurship. While taking Management 241: Knowledge for Social Impact Knowledge with Katherine Klein has taught me that the term gets thrown around a bit too lightly, I’ve come to a more rigid personal definition: making non-profitable decisions – those that wouldn’t happen in the simple tendencies of capital – in an effort to help people.

If I can build something, be involved in this very capitalist activity, and then make decisions that wouldn't necessarily lead to my profit or the company's profit and be for the good of my users or the good of my community, that would in some sense, be a social impact. Some things I’d love to do are underprice our subscription costs, offer financial aid to Title I schools, and build other ways to stand for teachers.

This, I think, could offset some of the negative externalities that come with building tech in a capitalist system. 

Let me shift to a hard truth. One I am sure you have considered. Almost all start-ups fail. That’s the nature of capitalism as practiced by VCs. Leaving aside questions of social good, what about your individual good? How do you think about the fact that success, at least as it is usually defined, is very unlikely?

I'm just gonna keep on going back and do it again. Yes. Most startups fail. But if I keep on trying, I think due to the law of large numbers eventually, maybe I'll hit. I don't know how well that logic works, but to some degree, I think that's true. Even just to get into Y combinator, you have to repeat. Their acceptance rate skyrockets for people who applied the second time. And even then, getting into Y-C is very far from building a successful company.

That's why this summer, I am working for another startup. If this one doesn't work, you know, I just keep on plugging away and learning. This is not my first time trying to start a company. But I think I've learned a ton each time so hopefully, one time I'll learn enough to to strike it. This is my third.

So give me the elevator pitch for a previous one.

The first one just stopped only because of the time where I was in my life. And I’ve thought about picking it up again. But I went for SwiftScore and LLMs because I think that's a more powerful idea. The first one was originally called Fox Academy because my school is called Fox Lane, and essentially it was a school-specific version of Khan Academy. 

We'd have students who are preparing for their exams shoot a video. They’d prop their phone up and take video of themselves filling out a worksheet or reviewing content. And then they would upload that to our platform. Then next year, students could watch those videos to prepare for the same test.

The concept was similar to Khan Academy. You know Khan Academy is awesome. I loved it. I used it. But it was missing information I needed. That's because teachers, the school, have their unique curriculum. So our platform had students teach other students, those who came up after them. 

Do you have something in your pocket if SwiftScore doesn’t make it? Do you know what you will try next?  Or, is it more like you've got ideas that come and go, and you will just have to see what happens?

I think it's the latter. Right now my head is just all in with SwiftScore. I have to think this is the next big thing, you know. And if it doesn't hit, I'll go back to the drawing board.

I have to circle back to taking a class with Jon Zimmerman. I don't know if you know this, but he wrote a Washington Post editorial last August, saying that he wasn't going to going to have a policy about AI in his class. He made the case that not using it would provide a better education and left it at that. As an AI edtech founder, I am curious about your experience in the class

This class is awesome. I don't think our views are incompatible because a class like Professor Zimmermann's is nothing I want tech to touch, you know, or I don't think it should touch. I don't see how it'd be helpful in this very intimate setting. He's a teacher who cares so much about his students. And when he grades, it's not just like, “Oh, I'm checking a box. This one’s doing great.” 

This class is focused on critical thought and oriented around writing essays. He connects with students and understands what they're thinking.  I think classes like that, I don't think they need ChatGPT. If a student in the class said, “Oh, I use ChatGPT to help me, to help read the assignments, and then ask it some questions. Whatever. I think that's fine. And if they miss out because they're using ChatGPT, then that's on them, right? 

I think there will always be classes that SwiftScore shouldn't touch and this is one of them. Now, if have a 50-person class and a high school teacher who goes home and spends Sunday grading assignments that are all right or wrong answers, then I think that's a bit different.

This has been a great conversation. Discovering you were taking history of education was such a nice surprise. Anything you wanted me to ask and I didn’t?

Not really the only thing I was. I was curious what you thought about Less Wrong and rationalist thought. 

When I talk philosophy with students in engineering, that’s almost always what they know. But it just seems so weak, so basic. I am convinced there is an opportunity to get the people building AI into something stronger. So much social theory is grounded in Marx or Nietzsche, so either waiting for a revolution that won’t happen or deeply skeptical of everything. I write and think about a group of philosophers called pragmatists, William James. Anna Julia Cooper. John Dewey. 

We've read a good amount of Dewey in Professor Zimmerman’s class.

I can’t tell you how happy that makes me. 

He's awesome. I agree.

So when's the bus trip to Boston to Y Combinator?

The bus trip is tomorrow.

𝑨𝑰 𝑳𝒐𝒈 © 2024 by Rob Nelson is licensed under CC BY-SA 4.0.

---

One of the few things I am certain about when it comes to generative AI is that the people graduating college in the next few years will be the ones to decide our AI future. So, I am considering making What One Student Thinks a semi-regular feature of 𝐀𝐈 𝐋𝐨𝐠. I would interview students, like Will, who have founded educational technology companies or are developing products that use generative AI for learning. Let me know what you think in the comments.