An Applied Mathematician, a Newsroom, and a Whole Lot of Data
Moonshot: Engineers Achieving the Impossible | Episode 3
What happens when a mathematician who studies biological systems spends a sabbatical in a newsroom? Chris Wiggins, an associate professor of applied mathematics at Columbia, joins host Mike Massimino to talk about predicting what readers will do, the media industry’s big shift in the digital age, and why ambitious ideas often start as small experiments with the right collaborators. Appointed in 2014 as the first chief data scientist at The New York Times, Wiggins now leads machine learning and AI science at CNN. Looking ahead, he considers journalism’s next transition and how the media can navigate disinformation and AI.
Episode Highlights
00:00 – Journalism’s digital transformation
18:18 – Joining The New York Times as its first chief data scientist during a period of uncertainty
34:51 – Proving readers would pay for content
42:32 – Using math to make sense of complexity
48:27 – The rapid growth of AI, changes in society
59:54 – Why being a researcher creates opportunities for lifelong learning
Transcript
Chris Wiggins: Journalism has become not something that was delivered in the form of a dead tree or in cable news, but in the palm of your hand. People are getting their information in very different ways and each of the ways that they get that information is itself a communication, just a constant communication between their app and the company or the servers, the computers that are trying to deliver that content to people and trying to learn from the way people engage, like what are people interested in and how do we create a product that's compelling?
Mike Massimino: I'm Mike Massimino, former NASA astronaut and Columbia engineering professor, and this is Moonshot.
In 2002, I launched into space for the first time aboard Space Shuttle Columbia to service the Hubble Space Telescope. People always ask if I was scared.
Honestly? You're heading into uncharted territory. You can prepare all you want, but you don't know what problem will show up or what you'll have to do when it does. So it comes down to this: trust your training, trust your gear, trust your team, do that, and you can attempt things nobody's done before.
Today, my guest is Chris Wiggins. In 2013, journalism was collapsing and Chris, an associate professor of applied mathematics at Columbia, made a strange move. He joined the New York Times as its first ever chief data scientist, back when nobody there had even used the phrase machine learning.
Nobody even knew at the time if readers would ever pay for news online. He built a team from scratch and helped the Times flourish while the rest of the industry shrank. Now he leads machine learning and AI science at CNN, doing a harder version of that same fight, this time against generative AI.
So Chris, welcome to the show. Thank you for coming. I know you're a busy guy. We work in the same building.
Chris Wiggins: Yes.
Mike Massimino: And, but we don't necessarily run into each other a lot. We're on the second floor, but you, so tell me, how long have you been at Columbia? And where, what are you up to?
Where do I find you over there? Because we are on the same floor, but I guess not really. You've changed your office lately.
Chris Wiggins: It's true. So from 2001 until 2014, so 12 years ago, I left the Times.
Mike Massimino: It's like geological time. Yes.
Chris Wiggins: Negative timescale is very recent. Yeah, 2001 to 2014, I was down yonder on the second floor in a room that not only didn't have a window, but worse, you could tell that there used to be a window before they built CEPSR, like what they put up shortly after, like 1994 or so. So you could tell from the cut of the cinderblock behind me that there used to be a window there, and no longer, to make room for CEPSR, which I'm very happy we have.
But then in 2014, I moved to what used to be the library. So when I was young, when you and I were young, there was an engineering library there, right on the fourth floor as you walk in from campus level. And that is now the Data Science Institute.
So in 2014, I ascended from the second floor to fourth floor, which is a lot of fun because the people around me are, this guy is CS and stats, and this person is CS. Before that, IEOR. Across the hall are people from IEOR and a bunch of other departments.
So it's an extremely interdisciplinary place, which is, you know, academia doesn't always foster people talking to each other across disciplines or schools or departments. So that's been nice.
Mike Massimino: The Data Science Institute. It sounds like this was a good idea. It worked out, put everybody together.
Chris Wiggins: Yeah, I thought it was a good idea.
Mike Massimino: Okay.
Chris Wiggins: The Data Science Institute.
Mike Massimino: Yeah.
Chris Wiggins: Mayor Bloomberg thought so too.
Mike Massimino: Really?
Chris Wiggins: Yeah, and the initial funding for this, you know, space, this beautiful building we're in, came from a competition from Mayor Bloomberg when he was still mayor to create a new Applied Sciences Initiative. You ever been to Cornell Tech on Roosevelt Avenue?
Mike Massimino: Yeah, I remember that one. Yeah, yeah. Yeah, I remember when that was going on. That was, what, 12 years ago or so, there was a competition for that one. Oh, okay.
Chris Wiggins: And Dean Don Goldfarb, who you mentioned earlier, was the dean at the time for the ribbon cutting in this very building, down in Joe's Coffee, to announce the creation of the Data Science Institute.
Mike Massimino: Wow, okay. So in academic circles.
Chris Wiggins: I got myself an office out of that by being there.
Mike Massimino: Nice, okay, all right. All right, so that's where your office is, where you're living now.
Chris Wiggins: Yeah, it's down on fourth floor Mudd, surrounded by a bunch of other people thinking about data and how to make sense of the world through data.
Mike Massimino: And it's a good collaboration. It actually worked out, because it sounds, in theory, that would be a good idea, get people together who have these similar interests from different fields. And it seems like, in this case, it's a success story.
Chris Wiggins: If you pick well, yes.
Mike Massimino: If you pick well, yeah. Pick the right people, you mean?
Chris Wiggins: Well, also a field, you know? Like, the dynamics of fields are really interesting. And when you're in the making of a new kind of science, to misquote, you know, James Gleick's book from the 80s, Chaos: The Making of a New Science?
Mike Massimino: Okay, so, man.
Chris Wiggins: Well, like, when I was in high school, this book came out about chaos and the making of a new kind of science. And it really kind of struck my imagination. How do you know when you're creating a new kind of science, as opposed to something that was just a calculation that was kind of cool, or maybe it's not going to go anywhere, or something like that?
So, you know, Data Science Institute was a bet. You know, Dean Mary Boyce, later, and Don Goldfarb, you know, they took a bet that there was really a coherent craft there and trying to make sense of the world through data, which wasn't real clear. And, you know, I'm old enough that I had seen, you know, nanoscience, systems biology.
You know, a lot of times when people suddenly realized that there was a field there that was not captured in traditional disciplines. You know, like, we have a mathematics building here. You know it's a mathematics building because they chiseled in stone on the side of the building, Mathematics. And you pretty much know that they're going to study mathematics there. And everybody pretty much knows what mathematics is. But then you get other fields like statistics or data science or artificial intelligence.
And the names of those things mean different things in different years, in different communities, even statistics, which we think, well, it's meant the same thing forever. You know, you can look back and see how what people meant by statistics was really different 200 years ago. And certainly what people are doing in the name of machine learning was very different in the eighties, when you and I were undergraduates from today.
Mike Massimino: You've got an interesting background. Your undergrad here was in…
Chris Wiggins: Majored in physics, double minored in math and religion. Yes.
Mike Massimino: Okay, so very ambitious student, interesting combination of things. And then your theoretical physics PhD. And now you work in all sorts of things, advising the newsrooms and cable networks on AI.
How would you describe to the audience here, what you do? Is it possible even to describe what is it that you do nowadays? What has your interest nowadays? What do you do now?
Chris Wiggins: You know, I'm still interested in how people make sense of the world through data, which is very hard. Sometimes the data are natural sciences data created in an experiment. Sometimes, you know, in the case of genomics, really it was sort of created by centuries or millennia of, you know, slow change that you can track.
In digital products, the data are often created by the way people interact with your digital product. And that's been a real transformation for journalism as journalism has become not something that was delivered in the form of a dead tree or in cable news, but in the palm of your hand, right? Like people are getting their information in very different ways.
And each of the ways that they get that information is itself a communication, just a constant communication between their app and the company or the servers, the computers that are trying to deliver that content to people and trying to learn from the way people engage, like what are people interested in? And how do we create a product that's compelling? Particularly, again, like, if you think about centuries of journalism where people spread out a dead tree on their paper and engage with it in long form in a focused way, that's very different than people who are trying to understand what's going on in the world, you know, while they're waiting for the light to change, while they're waiting to cross the street or something.
There's many different needs and people might be going there to be entertained, surprised, delighted, informed, to figure out what to cook that night, to dinner, to play a game. Like the whole experience of journalism is very different now than it has been for centuries. And a large part of that is this direct relationship between the company that's producing the news and the individual consumers who are telling the companies what they're interested in, what they want to know, how they want to use it, if the companies can listen.
And that's really what this digital experience opens up, opens up the possibility that these companies can hear from their audience and try to figure out, like, what are people interested in? What do they want to know more about? How do they like engaging with this story?
Mike Massimino: One of the classes you're teaching now is, it's a new class with a Poli Sci professor, which is a very interesting pairing. Can you tell us a little bit about that? How did that happen?
Chris Wiggins: So it's not the norm in academia for you to teach a class with somebody in a different department even, let alone with a different school. And I would say that's one of the best things that Columbia's done over the last couple of years, couple of decades really, is to encourage faculty to try to innovate by co-teaching a class with somebody from a different school. And it's very difficult to do well, and it's very educational, and I think it can be a lot of fun if you're brave enough to go from a field where you've become basically a black belt in your field, and then you go to another field where you are like a white belt, like you don't know anything.
So that's been very educational for me. So I co-taught a class with a historian, and I don't think I've taken a history class since ninth grade. And then I taught this class with a political scientist, and I've never taken a class in political science or economics for that matter.
So it was very, very educational to learn the way economists and quantitative sociologists and political scientists think about the way we make sense of the world through data. And then that was really her expertise. And then I was teaching about how companies do think about how to optimize engagement with content or how to convince people to do a thing, and the thing might be just to click, or it might be to purchase in the case of digital marketing.
In the case of politics, it's to convince people that a candidate or a position might be worthy of their eventual vote, which is, again, a very difficult thing to measure, and it's a very difficult thing to know if it works, which is part of what the class is about.
Mike Massimino: Is it trying to get political messaging? Like a politician, if someone wants to run for office, would they use some of the information you've spoken about to figure out whether…
Chris Wiggins: Not only would they, but they do. Absolutely do. It's been true for a long time. So there's this nice book, Making the President 1960, about how politics and marketing have long talked to each other.
Actually, one of the things we wrote about in my book with Matt Jones was Edward Bernays, who's sometimes called the father of public relations. And in early days, he sort of organized this event with a bunch of women smoking cigarettes during the Easter parade, during an Easter parade, and that was sort of an organized PR campaign to convince people that it's okay for women to smoke. And of course, Bernays also went on to work with Coolidge and other people trying to convince the electorate that those people were cool, which apparently was hard to do in the case of Coolidge.
So there's a longstanding relationship between marketing and political messaging, which has changed qualitatively in the digital age. So it's been true for decades that people have been realizing that the same computational infrastructure, the same algorithms are the ones you would be using, whether you're trying to convince people to like a particular brand of soap or to like a candidate. Like the mathematics doesn't know, and the mathematics is the same mathematics, trying to figure out how you send different content to different people, figure out how they engage, and then amplify the thing that seems to be working.
Mike Massimino: Is there a time where this became a whole different game as far as using the data? Was it based on the computing power or people looking at this? In other words, before it seems like they might've been these creative ideas in marketing, but now they're basing it more, and I think it's what you're saying, is that it's more like in real numbers.
Chris Wiggins: They can, and they can turn it into optimization. And to be clear, mathematical techniques like this are great for climbing one hill, but it won't help you figure out, actually, there's some other hill over there that we should be climbing. And there's still a lot of domain expertise in figuring out, you know what, this is a different type of messaging that would totally resonate.
And of course, there's other things like what the people actually, what the candidates themselves actually believe or what the product actually is. I don't know that there was one step change. I would say the step change is when people suddenly become aware of something, which gets back to journalism and people's awareness of things.
So often what happens is that a thing is changing very slowly, often in industry first, where there's room to explore and to figure out what seems to be working and to invest more in things. And then suddenly people become aware of it. But the thing that's a big change is not so much that the tech changes, just suddenly somebody becomes very aware of it.
This is true for large language models too. So November of 2022, people launched a product called ChatGPT, but language models had been around before. Predicting the next word had been around since Claude Shannon in 1948.
I mean, this is an old idea that you can take a corpus of text and use it to predict the next word, it's an old idea. But it's a product innovation and a change in people's mindset. So similarly with political and marketing messaging and the role of computers therein.
I mean, people have been doing that for a long time, but it's only a couple of shock events that suddenly grabs everybody's attention. And they're like, holy moly, we didn't realize you could use computers to optimize political messaging in the same way you were optimizing everything else.
Mike Massimino: Interesting. And it's also like in the world of sports. I recently watched Moneyball. I hadn't seen the movie in a few years.
Chris Wiggins: No, it's a good one. Well, I know the book a little bit. I know the movie a little bit, but I know the book well.
Mike Massimino: Yeah, and they're using data to make decisions. The general manager of the Houston Astros, years ago, had a sign on his door, he came in and was applying all these different techniques. And a sign on his door was, "In God I trust, everyone else must bring data." So that was, I don't know if that was his quote, but I remember seeing that.
Chris Wiggins: Not originally. I'll tell you though, Moneyball is a great tool for explaining to people how you can make sense of things through data that you wouldn't think of as being a data-driven problem. And that could be a company trying to convince you which movies to watch or which wine to buy, the company trying to convince you which books to read.
And Moneyball is a good example because Michael Lewis, the author of the book, does a great job illustrating how there was a whole culture and a whole lore for the way you should understand what makes a baseball player good. And then there's a totally different way of doing it, looking at data. That said, one of the things that you can only do so much in a book is explain, actually, it's very subjective.
What is the number that quantifies a baseball player being good? There's a whole bunch of different ways to do it. And there's a whole tradition of amateur baseball researchers, Society for American Baseball Researchers. I don't know if you know that.
Mike Massimino: No, no, no.
Chris Wiggins: Okay, so long before the book, there already was a whole bunch of people who for decades were doing this kind of craft, just taking a bunch of data. And then it's not clear because you don't have F = ma. You don't have some sort of basic physics that's supposed to drive things.
And so there's actually a lot of subjectivity even though there's data involved. And that, I think, is one of the most exciting things about data science is trying to remind people, like, actually, just because there's data involved doesn't mean that there's one answer. There's still going to be a bunch of conversation about like, wow, okay, how do we make sense of the world through data? Baseball, you know, perfectly fine example.
Mike Massimino: Yeah, and is that where the people come in with the expertise to analyze it?
Chris Wiggins: Yeah, there's a lot of domain expertise and there's a lot of people asking themselves, what does success mean, right? And it's very rare that success is, you know, reducible to a number, to one particular number. A lot of times success is something that we don't even attempt to quantify at all.
If you think about journalism and impact, for example, a lot of that is the stuff that cannot be counted, right? Another adage like yours is, not everything that counts can be counted and not everything that can be counted counts. So a lot of the things that are most valuable in journalism are not things that you easily turn into numbers, right?
So that's part of the graph, just like Moneyball is to figure out, okay, we can quantify stuff, so what? Right, and this happened in biology too with the advance of genomics in the 1990s and thereafter. We can actually measure the sequence of freely living organisms.
This was an innovation from 1995. Okay, great, so now we can sequence a whole bunch of mice and men and chickens and C. elegans and Drosophila and can sequence a whole mess of people that are related to each other, now what, right? And that I think was one of the most exciting things for me as then a graduate student is there's so much data and yet not real, like we don't have F = ma or some sort of basic model we're supposed to compare it to.
So that's a statistical and societal challenge. How are we going to make sense of data? How are we going to answer questions we care about using this abundant corpus of data?
Mike Massimino: All right, let's go a little bit back to some of the origins of this. We talked about journalism. Your sabbatical that you had in 2013, right?
That was most of the newsrooms were watching their industry collapse apparently. It was, they were cutting staff, they were, because the digital age had come in. But the New York Times made a different bet and they wanted you to come, tell me how this happened.
Did the New York Times look at things maybe a little bit differently or did they not? Or did you convince them? What happened there when you went, when you took that sabbatical and went over, how did that happen and how did it change things?
Chris Wiggins: Yeah, it's a good story. It's a weird story. It was a weird year. So I don't want to take too much credit for the transition. I mean, the New York Times lost a digital paywall in 2011.
Mike Massimino: Okay.
Chris Wiggins: And it is true that when they lost a digital paywall, a whole mess of smart people wrote articles saying that was a terrible idea. And by 2013, when I decided I needed to take a sabbatical, it still wasn't clear if there was a repeatable, sustainable model for journalism anymore, right? Very exciting.
So in 2013, I decided to take a sabbatical and I went and asked Mark Hansen over in the journalism school for advice about what to do for my sabbatical. And Mark has been on the faculty since around 2010 or so, if I remember correctly, got his PhD in computational statistics. And so he's a very, like many great Columbia professors, he knows no disciplinary boundaries.
Mike Massimino: He's in journalism, but he had this, yeah, okay.
Chris Wiggins: So his advice to me was, don't just go to one of the big tech companies because that would be like your day job. You should go to the New York Times. It would be weird. And I did, and it was. So I showed up at the New York Times and I met people from the newsroom, the business side, R&D, lots of different people. And it seemed like the place where I really could have a lot of impact was the team that had just started collecting data about the way people interact with the product.
So if you think about a company like LinkedIn or Netflix or Facebook, the first thing they do is think about is how are we going to log all the interactions, right? We're going to become a, we're a data-driven company. If you think about a company that's been around since 1851, right, that's not necessarily the first thing they're thinking about.
And arguably, they shouldn't be thinking about that, right? If you want to earn the trust of the readers, you should think about being great at the craft of journalism. So in 2013, they had just started keeping track of all of the digital events when people interact with the product.
And at the time I had been working in machine learning, applied to biology, and I had a real confidence that methods of machine learning could reveal things that are of interest to biologists. And so that's how I spent my sabbatical was putting my head down and writing some really bad code to try to figure out, can you predict things that the leaders of the company care about from the data that they already have?
Mike Massimino: And it kind of, it worked, yeah? Were they able to understand what was happening?
Chris Wiggins: Understand and optimize. So one thing about academia in the real world is often in academia, you want to understand something and you want to know what's true. And in applied fields, maybe engineering, maybe industry, maybe medicine, it's not just that you want to know what's going to happen in the absence of some sort of treatment.
You actually want to know what is the treatment that's going to drive some sort of outcome you want, right? Like if you're a doctor, it's not enough just to predict who's going to get sick and who's going to get healthy. You actually want to figure out what treatment should I give that's going to make people healthy.
And in industry or in engineering, control, right? Like if you think about control, it's not just about predicting where the thing's going to go. It's trying to figure out, okay, how do I learn how to drive the thing in such a way that it's going to go towards where I want to go?
That's absolutely true in industry. So it's not just that you can use data to understand, for example, who's going to cancel their subscription, who's going to become a subscriber. It's that you want to use data to figure out like what is the thing that I could do to get somebody not to cancel their subscription or to convince somebody now is the right time to become a subscriber.
It's a difference between a type of analysis sometimes called predictive and prescriptive. It's not just predicting what's going to happen in the absence of treatment. It's trying to prescribe what is the treatment that's going to drive your outcome.
And one of the ways it's different from many of the things we do in research is the only way to do that is to intervene with the system. Otherwise, you could never tell if you're doing correlation or causation. So it's a great thing to do in engineering or in health or in industry.
But when you think about a lot of the things that I had done heretofore in the sciences, it wasn't so much about control or trying to decide the optimal treatment. It was really just trying to figure out can you figure out how does C. elegans work? Or like how does yeast regulate its genetic network or something like that? Those are things where you're asking how it works questions and not questions about how do I drive it to a particular outcome.
Mike Massimino: With the Times, was there something at that point, you said they went to the paywall in 2011.But did they approach you? Did you approach them? I mean, did they realize they had to do something?
Chris Wiggins: I approached them.
With the help of Mark Hansen, my colleague, who made the right intro to somebody in the newsroom, then I started meeting other people and saying, I'm looking for a place to do a sabbatical. Is there a place here where I could contribute?
Mike Massimino: And what did they say, yeah, we need help? Or did they not know that they needed help? Or what was that like?
Chris Wiggins: It's not easy for people in industry to say too much until they're actually in the door about what they need. There were clearly ways that I could be helpful to, for example, R&D and think about new research ideas. There were some ways I could be helpful to the newsroom, but I want to emphasize that like computer assisted reporting is an old field.
So the things that I could do in computer assisted reporting were not particularly, would not have been particularly transformative. R&D is a great field of data journalism and great people who do that. The thing where I felt like I could really make a difference was on the business side, as they would say, which is they're gathering all this data for the first time about the way people interact with the product.
They've made this big bet that the future of the company is not in advertising, but as a digital subscription service. How do they figure out what do we do to get people to become digital subscribers and to stay digital subscribers? That seemed more clear to me, that that was the place where I conjectured I could have the most influence.
Mike Massimino: Was that the first thing that you built for them; was something to do that?
Chris Wiggins: So the very first day, I remember sending an email to my then-boss saying, what we really need to do is something like control theory, getting back to an earlier conversation we had. What you really want to do is to figure out, okay, we've got a digital product. The digital product can be optimized, but the software engineering is not in place to do that.
I mean, the data was barely there, let alone the connections between data and driving the digital product. And so instead, the first thing we should do is just build a churn model. So building a churn model is, if you take a quantitative marketing class kind of thing, anybody would tell you like, just watch people use the product and predict which people are going to stop using the product.
Mike Massimino: And you call that what, you said?
Chris Wiggins: It's called a churn model in industry. And it's, I now know it's a pretty standard thing to do. For me coming from biology, I was like, okay, well, I know how to predict which genes are going to go up and which genes are going to go down based on the regulatory sequence, based on the regulatory state of transcription factors, which are a type of protein that drive things. Let's just try to do the same thing here.
Can I just predict some binary outcome from a bunch of things I can observe now? So that was the first problem I worked on was like, can you just figure out what features seem to correlate with somebody who's going to stop using the product? Which was about all I could do given the state of the software engineering at the time.
So then you need to build out the connections, the connections of the gears, so that the data that you draw and store and collect and measure is connected to actual digital decisions. And it could be blue button versus a red button, or it could be something like how much access does somebody get to the journalism before you ask them to become a subscriber? But either way, you really need to connect those gears. That is an engineering task. They call it software engineering. Everything that can go wrong can.
So when I first heard the term software engineering, I was like, is there really an engineering there? But having watched a bunch of things go wrong now, I feel like I can understand like, okay, now I understand why this is called software engineering.
Mike Massimino: And so you're able to look for the data that will tell you a person is likely to unsubscribe.
Chris Wiggins: Yes.
Mike Massimino: Right? And what do you do about it?
Chris Wiggins: Exactly.
Mike Massimino: You found them. Okay.
Chris Wiggins: So the truth is, again, getting back to my analogy with the way doctors work, if you think about like the COVID vaccine or something like that, they didn't just say, okay, give the COVID vaccine to everybody. They actually did a randomized control trial and gave different vaccines to different people at random. And that idea, which is probably a hundred years old in statistics, is still really powerful for figuring out how the world works.
You have to actually interact with the world in order to know what treatments are going to drive some particular outcome. So knowing which people are at risk, it's kind of interesting. Knowing which are the risky behaviors, kind of interesting.
But knowing what to do about it is really what you care about. If you are an engineer or a doctor, or you work in industry, that sort of prescriptive analytics is really more powerful than predictive analytics. But the only way to make that work is if you have a way to interact with people.
In medicine, that's a randomized control trial. In industry, that's an A/B test or some other sort of randomization. And that type of interacting with the world is crucial for figuring out what's the best treatment.
Mike Massimino: So it could be, and I guess it's different for each person, right? So it could be, we need more, I'm coming up with an idea.
Chris Wiggins: I see where your mind's going. Your mind's going like, we know this thing about a person, so what should we do? It's the opposite.
The opposite is, we have to figure out first, what is it that we want people to do? And then figure out, well, what are the levers we're willing to pull in order to drive that result? And then to pull those levers at random, and then pull those levers not at random.
That is basically the algorithm of all causal control, all stochastic optimization, we would say in computer science or statistics. That is one playbook, is you have to figure out how do we quantify what it means for something to be good? What are the levers we're willing to pull in order to get that outcome?
Pull those levers at random, and then pull those levers not at random. That's it. That's the playbook. And the thing that it flips, is that there's only so much intuition you can get from looking at dataset until you've actually interacted with the world. But once you've interacted with the world, then you sort of know everything about that set of interactions, which is why it is so useful to explore.
That's why it's so useful to try everything, is to like, until you've tried the thing, you just, you really don't know what's going to happen.
Mike Massimino: I'm just thinking, like the sports page, say if you're a person that reads the sports page and you want better coverage, the Times now, they purchase, I think the athletic at some point, right? Is that, I guess that was to drive their, keep people reading their paper, or is that a, oh, it was a, I'm just trying to think of an example that.
Chris Wiggins: That's a good example of a different hill. Right, so, you know, you call up a statistician when you're on a hill and you want to figure out how do I optimize that hill? But a big bet like that, that is something for which data is just not going to help you much.
But it's a good example about how when you're, you know, trying to make decisions about how to interact with the world. There are times that you can use data to figure out how to optimize the thing you're doing, but there's still a lot of important room for a human being to say, you know what? There's a different hill over there.
We're going to go climb that hill. And sometimes you do it in a way that's a small exploration and then sometimes you just make a big bet. Just like we're going to make a big bet that by, in this case.
Mike Massimino: That's a big bet, yeah.
Chris Wiggins: Buying that company, we're going to get a whole bunch of subscribers who probably weren't reading us before, right? If we establish a brand in this new area and you know, you can, there's some data work to be done there in the sense of, you know, consulting and you look at other, you know, comparable companies, but it is not a stick. It's not that playbook, right?
That playbook of randomly doing things. Like when they were, you know, trialing the COVID vaccine and they had to give random shots to random people. That's different type of strategy than the strategy of like, you know what? Data is not going to help us here. We're just going to make a big bet. That's the right thing to do.
Mike Massimino: Yeah, in my case, I had a subscription to each. The Times and the Athletic. And then it took me a little while to figure out, I don't need both.
And I canceled the Athletic and I still get the Athletic included. So it worked out. To do these things, when you went there, were you kind of the first person doing these things? And then you needed to build a team at some point?
Chris Wiggins: Yeah.
Mike Massimino: And then was it an issue to try to get them convinced that they needed to build a team to do this and not just one person?
Chris Wiggins: It's a big enough company that I can't say that I know exactly all the conversations that were happening in rooms I wasn't in, but somebody was willing to let me show up there to do a sabbatical. And after a couple of months, I said to them, you know, we should make a machine learning team. And most of the reaction that I got was, what is machine learning?
And I said, okay, we should make a data science group. And they said, that sounds okay. Let's build up a data science effort. So I helped them hire some people. There was not a lot of, to my knowledge, when I was there in the summer of 2013, I asked around, I was not able to find people who knew machine learning. But we were able to cobble together.
I found my doers, to use a phrase, which is I found other people who were willing to try something weird and to try to put together the data to build predictive models, to use those predictive models to drive insights, to convince people that those predictive models should become prescriptive models, meaning we should actually drive experiments and then use that, the data produced by those experiments to improve the product and improve the experience for people. So that took a while.
Mike Massimino: Where'd you find these people? Did you come back here to Columbia or were they other places?
Chris Wiggins: Oh, no, no, these are all people in the building.
Mike Massimino: People in the building, okay, great.
Chris Wiggins: One guy was a guy who I had met when he interviewed to be a master's student here at Columbia a year earlier. I had tried to convince him to come do a master's here. Ended up not doing that and ended up going to the New York Times. One was a summer intern who was doing his undergraduate degree in computer science. And then eventually I started helping them hire people. So the first person I hired was a computer science PhD from Brown and then hired a bunch of people, lots of Columbia alumni in point of fact. And that group grew to be a bunch of smart and nice people.
Mike Massimino: Do you think they're interested in both working in journalism and in data? It seems like you have a lot of, you're able to apply your interest in mathematics and data science to a variety of things. Do you think it's a way for people who are maybe interested in journalism but aren't necessarily inclined to be a journalistic writer that they have a passion for journalism and for data? Is it like, it allows you, or is it more the data and they're applying it to this one area?
Chris Wiggins: So you're getting at something really interesting there in the question, which is you can have the passion for journalism without being good at it at all.
Mike Massimino: Yeah. It's like in the sports world too, yeah.
Chris Wiggins: That's right. Or in the office or something. So I do think that there are people who go work at a place like the Times and there are many places like the Times where there's a mission and they go there and they could put their talents other places, but they think I want to use my talents at a place where at the end of the day I feel like I'm advancing some mission that matters.
That's different than hiring people who think, oh, I'm good at journalism. Because I assure you, I'm not good at journalism. The phrase sometimes news people use is, do you have ink in the veins? And I do not have ink in the veins. I'm not a journalist. There's things that I'm good at and that is not one of them.
So I think by and large, we hired a group of people who are really good at making sense of the world through data and developing digital product and collaborating with each other. Not so much people who felt like they were both going to be good at journalism and at data. There's a few people who could do it.
Again, Mark Hansen is a good example. Somebody who actually has bylines at the New York Times, writes stories, but also is obviously very good at data. Most of the people I was working with at the New York Times or now at CNN are people who enjoy making sense of the world through data, but not necessarily good at journalism, but like that they're working at a place that's trying to advance journalism.
Mike Massimino: You said earlier how in 2011 they started the paywall and that people were very skeptical. No one's going to pay for news. What did you think about that statement?
Do you think that, yeah, there's some truth to that? Or you're like, no, that's totally wrong. People will be willing to pay for news under the right circumstances. Did you see something different than what those experts saw?
Chris Wiggins: I think when I started, I just thought it was a really great technical challenge for a company that needed to be saved, like a really important craft, the craft of journalism. And I really wanted to do what I could to help prove out that there exists a repeatable, scalable business model for journalism in 2013, which in 2013 was absolutely not clear. Part of this relates to something that's not scientific, though.
It's norms. Like, are people willing to pay for content? 2011, it was not so clear, but if you think about today, people are subscribing to Netflix and Spotify and Headspace and other digital subscription services.
In retrospect, it doesn't seem as big a bet. But at the time, there's a bunch of people writing, information wants to be free, as people often say. And they very rarely quote the full paragraph, which actually says that sometimes information wants to be very expensive.
Information is valuable. And so it was a bet that people would be willing to pay for something that many people felt like they should get for free. But there's no physics there. There's no law. It's just a question of people's norms. Are they willing to do it? Can you convince people that it's a good idea to pay for digital subscription? And clearly, a bunch of people are.
Mike Massimino: I'm one of them. I think it's worth it. Absolutely. You said, you can verify this, that the biggest lesson from your Time's years wasn't really about the math. It was about understanding people and complementary skills.
Chris Wiggins: Yes.
Mike Massimino: Do you agree with that?
Chris Wiggins: Oh, yeah.
Mike Massimino: All right, so was there an actual moment where you can recall where someone with a completely different skill set saw something in the data that you couldn't see alone just by looking at the data?
Chris Wiggins: All the time. Constantly, I'm thinking about things where I'm looking at data like nuts. To me, I say nuts, and the other person says eureka.
That tension between nuts and eureka, it's there all the time, where you look at the data like, that looks weird. And then you show it to somebody else, you're like, oh yeah, I can totally understand.
Mike Massimino: Really?
Chris Wiggins: This is the aspect of user behavior, or software engineering, by the way. Because somebody knows the software well, they know, oh, okay, well here's the pathology that makes the data fall out that way that you wouldn't have expected. So yeah, if you find people who are smart and have a complementary skill set to you, you get to surprise and delight all the time, which is a good argument for finding people who are smart and complementary to you. You learn a lot.
Mike Massimino: So just different perspective on things is very useful, even in looking at the data.
Chris Wiggins: Yeah, or you think about it as a team, right? We can't all be place kickers, right? Like, if you want to get things done as a team, you need to make sure that you have a team where different people are good at complementary skills. So together, you can get a lot done.
Mike Massimino: You said earlier about, and it's something, a line from Richard Hamming: In science, not knowing if something will work is the whole point, right?
Chris Wiggins: Right.
Mike Massimino: How is that accepted, or how is that perceived, that notion by the folks in the newsroom?
Are they like, no, we need answers, or are they okay with, hey, we're going to see what happens here, and then adjust if needed?
Chris Wiggins: Right, that tension between science and engineering, it shows up in talking to people in the newsroom, but also for software engineers. So if you think about building a bridge, you're pointing at an excellent bridge out there, human beings, the GW right there, and it's still there.
Mike Massimino: The GW, George Washington Bridge. Still there.
Chris Wiggins: People have been building bridges for a long time, and you should not build a bridge unless you know what you're doing. But for something like creating an optimized digital paywall that makes new decisions every minute to every hour about who should see which content besides, or otherwise get the paywall, that is a thing that had never been done before. And so that's more like research and less like engineering.
It's more like, we're not really sure this is going to work, we're not really sure that there's enough signal in the data, we're not really sure that we can get the pipes to work in terms of software engineering, and we're not really sure whether or not people will dig it. It's possible that the market rejects this innovation entirely. So there's a lot of unknowns there.
So I think about that Hamming quote sometimes where you have to convince people, we're doing this thing, we're not really sure it's going to work, we have to think carefully about what is the smallest experiment we can do such that the result of that experiment will show us whether or not it's worth doing another experiment.
Mike Massimino: Okay.
Chris Wiggins: And that's the thing about big bets, like buying a company as opposed to putting data to work is how do you figure out what's the smallest thing you can do so that you can figure out from the result of what you did whether or not it's worth doing a bigger thing.
Mike Massimino: So the baby steps in there.
Chris Wiggins: Yeah. And a baby knows where they're eventually going to go, right, the baby is stepping and eventually that baby's going to walk just like people the baby sees. Often in science you don't actually exactly know all the rungs on a ladder.
You have this conjecture that there's a ladder, you kind of know what the top rungs of the ladder look like and then you need to figure out what is the first rung and the second rung such that if we climb these rungs we might figure out what the top of the ladder looks like and be able to get there. So it's definitely in between in science and engineering.
Mike Massimino: Do you think it's in some ways liberating to have that mindset, in other words, well, we don't know if we're 100% correct here, but that's okay and maybe it'll work and maybe it won't, let's give it a try. It kind of takes, maybe takes the pressure off a little bit, but we're taking a risk here, it may be okay, it may not.
Chris Wiggins: If you create a good community where risk is tolerated, then yes. Most of it, it goes back to what you just said in passing, you just said it's going to be okay. It takes a lot of confidence to go do something weird and new, right?
And so you've just got to believe that this thing we're going to do, it just might work. And if it works, it's going to be great. Let's go do it, it's going to be great.
I do feel like that's part of my value in a lot of different rooms that I'm in is just telling people like, it's going to be great. Let's try it.
Mike Massimino: Yeah, let's give it a try. But I always think, well, what's the worst that could happen?
Chris Wiggins: Well, if you set things up right, and again, that thing about the rung on the ladder and baby steps, which is you don't let babies try to learn their baby steps in high-risk situations. If you set things up right, you do the experiment in such a way that it is a small experiment that gives you a result that clarifies things. So that is a craft too, just like for a scientist who thinks about what's the toy problem that's going to help us get to the full research problem.
In engineering or industry, there's a craft to think about, okay, well, what is the smallest experiment we can do such that that experiment will nonetheless tell us whether or not it's worth doing another experiment? That's a real agile mindset for developing products.
Mike Massimino: Your path has been in many different places, physics, biology, your minor, again, was it religion?
Chris Wiggins: Double-minored in math and religion.
Mike Massimino: Double-minored in math and religion, and then machine learning at its beginnings, journalism, AI, and now you're AI at CNN. So there's a lot of stuff going on there. Has there been like a kind of a through line there?
Like, well, it's all these different things, but has there been this kind of one passion or one line that keeps you through all these things?
Chris Wiggins: Well, I think a through line is trying to make sense of complexity and trying to think clearly and often trying to think collectively with other people about trying to make sense of complexity. So I mentioned earlier this book that came out when I was in high school, Chaos: Making a New Science. That field, which really blossomed in the 1980s, was about how do we understand complexity in nature and are there simple mathematical methods that will help us understand complexity?
So I do think that is a through line is how do we use mathematics, which is like a great tool for thinking clearly, how do we use that as a tool for understanding the complexity of the world? Math is often about reducing things to something simple enough that you can think clearly about it, and the audacious act of an applied mathematician is to think that you can use mathematics as a tool for modeling the complexity of nature, people's behaviors, shape of the clouds, fluid dynamics, whatever it is, to think about that subject, which is complex, using the methods of mathematics. I think that's the through line.
Mike Massimino: So it's about the data. You love data science. Is that safe to say that? Was there something that got you interested in this in the first place? It was something like maybe when you were a kid or you were here at Columbia or something that happened that said, yeah, this is what I'm passionate about. This is what I'm meant to do. Did something like that happen? What drew you to this?
Chris Wiggins: I mean, it was a series of question marks and exclamation marks, meaning there was a series of moments where I was like, how could that possibly be true? And a series of moments where I said, holy moly, math works. We should do more of that.
So for example, I mentioned sequencing freely living organisms. So that was first done in 1995, and I can remember well the cover of Science Magazine in 1995, the cover of which was the genome of Haemophilus influenzae organized as a ring. So Haemophilus is a critter that had already been known about for a hundred years and people had understood that organism by putting it in a Petri dish and pouring things on it to see if it turns blue or not.
I mean, there's a whole way of understanding that organism and its behavior, but to have access to the entire genome of that critter, a freely living organism, its whole genome, raises all these questions, right? So that was sort of a question mark moment. What are we going to do with all these data and how are we going to understand a century of understanding this organism in light of this new encyclopedia worth or maybe phone book worth of ones and zeros or A's, T's, G's, and C's over and over again? How are we going to make sense of those two things together? So that was sort of a question mark moment. And then a bunch of exclamation mark moments like in my research showing that you can use methods from machine learning that would show whether a network had evolved under one or a different network evolution mechanism.
So that was a topic of great interest in biology in the early 2000s. We have networks in biology: transcriptional regulatory networks, protein-protein interaction networks. Can we understand something about how those networks came to be? What sort of evolutionary rules are consistent with the resulting topology? Or a paper where we showed that you could look at different viral genomes and tell whether it had been evolving in a bird or a pig, basically swine flu or avian flu, just by looking at a nascent virus which had presented in the human population. And to show that those methods work and you can make an accurate prediction on held-out data.
Those were all sort of exclamation mark moments where I was like, holy moly, math. This thing that had been developed by people who were in very applied situations like optical character recognition or something that doesn't have anything to do with the natural sciences, nonetheless are extremely powerful when applied to the natural sciences when you have enough data. Or in industry, that these methods work in industry and can make predictions about who's going to do what. Each one of those is like, wow, math is good. We should do more math.
Mike Massimino: And this happened when, you mentioned that article in '95.
Chris Wiggins: Yeah.
Mike Massimino: So you were a student here then? You were a grad student.
Chris Wiggins: It's true. I was sitting in the Burger King.
Mike Massimino: Really?
Chris Wiggins: When another graduate student walked in holding Science magazine like, this is going to change biology.
Mike Massimino: Really?
Chris Wiggins: And I was like, why? Why is this going to change biology?
Mike Massimino: Wow.
Chris Wiggins: And one of the best ways to figure out something you don't understand is to really engage the papers in that field and then to start doing research in it or to start teaching with the field, somehow to put yourself in the ring is a great way to actually figure out what is going on in that arena. So whether it's publishing papers or teaching a class with somebody, all those ways of challenging yourself by stepping into a ring where you don't know the rules is a great way to find out real quick how things work.
Mike Massimino: And so at that point already, you already had your degree here. You're already in grad school.
Chris Wiggins: I was in grad school.
Mike Massimino: That's what you could do. So it wasn't, you loved math by this point. And now it was something where it could really apply in a new area that could make a huge difference in the world.
Chris Wiggins: Yeah, my time here was a lot about using mathematics for understanding complexity. So I did research in that building on chaos and dynamical systems.
Mike Massimino: In Pupin, in our physics building.
Chris Wiggins: Oh yeah, oh yeah. Took some great classes in chaos with the late Ed Spiegel. And did a bunch of research at Lamont-Doherty on geophysical fluid dynamics and trying to understand how patterns that we see in the world could come about from simple partial differential equations. So a lot of my time at Columbia was thinking about complexity. Complexity in mathematics, right? Math, you know, somehow you might think of math as the supreme arena of the simple and the world. If you haven't noticed, kind of complex. So those moments where you can say something clear and mathematical about complex systems, that's really what drove my curiosity as an undergraduate.
Mike Massimino: So we're facing maybe a different version of the 2013 problem again, where now it's the enemy and the tool maybe are the same thing, AI. Is this a harder problem to try to get your grips around, AI as opposed to what it was with machine learning all those years ago?
Chris Wiggins: So the complexifier is that it's not just about the tech, right, it's a complex socio-technical system and it's wrapped up in economics, nation-states, right? Like AI has now penetrated a number of battles, right? Economic battles, nation-state battles, the way people think about what's true, what is and what ought to be.
AI is a part of all of those, in part because AI has become a very powerful technology itself because of investments in computation and data. And in part because of just trillions of dollars of investment just since 2022 in advancing the technology and advocating for the technology, right? Because some of that trillions of dollars is setting up a system in which that is to be assumed, right, you assume that you have these features available to you when you use these technologies.
And some of those dollars are being spent convincing you that you need to have these technologies as a part of your daily life, right, incessantly. And that's not about scientific advances, right? So that's a major investment that impacts society, put multiple trillions of dollars into anything, it's going to impact society, which are not about making better math, right?
And in many ways, the technological advance of getting an algorithm to predict the next word from previous data, again, the idea was right there in Claude Shannon's work in 1948, down at Bell Labs, right here, probably here in the Meatpacking District, but scaling it up in such a way that it has shocking effects, right? And again, that's subjective, whether or not you expect a computer to be able to do that is a statement about our own expectations. That has been really transformative.
But to get back to something you just said earlier, I can't remember what you said, did you say the enemy or the friend or something like that?
Mike Massimino: Yeah, the enemy and the tool are the same thing.
Chris Wiggins: Exactly. So to my mind, it gets back to Kranzberg's First Law of Technology from the 1980s. Kranzberg said, "Technology is neither good nor bad; nor is it neutral." If we think about technology as a thing that delivers a capability to you, sometimes divorced from any intent, right? I give this capability to you, you decide what your intent is, right? Then there's, that opens up a wide arena, right? Because the person who developed the capability, they may have had one intent, but by making that capability transportable, which is one of the things technology does, you open it up to people with a wide variety of intents.
Mike Massimino: Are you optimistic about AI? There's different viewpoints on it.
Chris Wiggins: I mean, I would say I'm cautiously optimistic about almost everything in my life. And I definitely think you won't get anything done if you don't have some optimism. So I would much rather stay optimistic.
Mike Massimino: So it's going to help us, you're thinking?
Chris Wiggins: It will help and hurt, right? So it'll remain neither good nor bad, nor neutral. It's going to deliver. It's going to deliver challenges and it's going to deliver benefits.
Mike Massimino: So disinformation, is there anything recently that's happened that makes you feel optimistic about it or alarmed where we're headed? I guess it's kind of the misuse of data or maybe putting things out there that aren't quite true.
Chris Wiggins: Yeah, I mean, in some ways disinformation has a very long history. So I'm not necessarily surprised by things. How does disinformation have impact?
Well, in part because there's some sort of duplicitousness of some content that's produced, either because it is real, it's presented as real when it's not, or it's presented as coming from one context and it's actually coming from some other context. And there's so many ways to do that. I don't know that there's a clear, single engineering tool to act against it.
Instead, it's really in part normative because if people are perceiving something that's false, but they're expecting it to be false, then that lessens its impact, at least its impact for representing something to be true. Clearly there's people who are happy to share things that they know are not false. It's not even plausible because of entertainment or not necessarily trying to persuade.
The only things that give me optimism is the ways that people eventually become inoculated to all sorts of different things. And they become used to seeing something that's synthetically generated, but they don't necessarily give it credence. They don't necessarily believe that it's credible.
So I do think that's the best we can do is to make sure everybody knows here's what the technology is capable of in order that people, when they're presented with some new piece of information, have the appropriate level of skepticism. Like how do you go about doing your research responsibly to understand is this thing as it is presented or not? And that will absolutely continue to be an arms race.
Same thing with plagiarism detection for large language models. Large language models are trained on natural language. So, and their job is to produce stuff that's indistinguishable from other naturally generated stuff. So that's another example of an arms race. I don't know that these arms races are solved by building better misinformation detectors or building better plagiarism detectors. I really think it's more about people becoming informed about what the capability of the technology is and hopefully designing systems in which you don't encourage people to be adversarial.
If we set up a system in which we encourage people to be duplicitous about say AI generated content represented as their own, for example, that's sort of a design failure. We've set up the system in a way that's incentivizing duplicitous behavior and information ecosystems. Hopefully we can design systems in which we don't incentivize people to create duplicitous information in part because people are not swayed.
People don't get this information in a way that causes them to make decisions they would not have if they would have had time and resources to do more research.
Mike Massimino: Do you think the situation's getting better or worse or is the same?
Chris Wiggins: As with any arms race, right? So constantly red is beating blue and blue is beating red. So there's a lot of motivation for people to create duplicitous content.
And as long as there's motivation to, then I'm sure that it's going to keep happening. Yeah, that's right. So I don't know that it's going to get, I don't know that the scales are going to tip in one direction, but I do think it will continue to be an arms race.
Mike Massimino: And you're at CNN now.
Chris Wiggins: Yeah.
Mike Massimino: Tell us what's going on over there. What are your goals over there? What are you working on there with CNN?
Chris Wiggins: So one way to think about it is what's the same and what's different from my engagements at the New York Times?
Mike Massimino: Yeah, okay.
Chris Wiggins: So one thing that's, and this is going to be really inside baseball. One thing that's really interesting is almost everybody that I'm talking to at CNN are people that I was talking to at New York Times. So like the CEO is the same CEO who was at New York Times.
Mike Massimino: You mean he's moved?
Chris Wiggins: The people who run, the person running data, is the person who was running data at the New York Times. A lot of the people doing product or growth or engineering are people that I had already known a decade ago from the New York Times. So that's one sort of inside baseball is actually, it's like a little bit of a high school reunion.
Mike Massimino: Yeah.
Chris Wiggins: And a lot of the problems are very similar. So the problem is, what do you do when you've had a business model, which has been the same for decades or centuries, it's worked just fine, and then technology has changed and you have to figure out, is there a new, repeatable, sustainable business model into this new world? In the case of the New York Times, it was largely print and advertising that was steadily going away. Print meaning you go buy a physical copy of the newspaper or you're a home subscriber, you subscribe to a copy of the paper. Or the revenue source of advertising, which steadily over the course of three decades has gone towards Google and Facebook or other digital advertising platforms. So the New York Times had this challenge to figure out how can we move to a direct-to-consumer model, which I would say, they pretty successfully did. And CNN has a very similar situation, only it's the cable bundle.
So people subscribing to cable was very, very good for cable companies for a long time. Now people get their information from a variety of options, the palm of their hand, streaming services, advertising-supported platforms. So a variety of ways that people engage with information.
And so a lot of the problems are similar in that there's a question, can you create a digital product that is sufficiently engaging that people will want to pay for it? And that's a similar challenge to the challenge New York Times faced 10 years ago. The thing that's new and weird is the linearity of television.
So people sometimes talk about linear content as being what you get when you sit down and you turn on a television and whatever content that's going to come next is the one content that's going to come next for everybody. And that's very different from when you read a newspaper and your eyes go from article to article, or you throw away section A and you go straight to section C and you're really interested in that. There's a non-linearity for that, that you explore the content as you want. Or when you're reading in the palm of your hand, you can click on a different article and you're exploring the content in any way you want. TV, on the other hand, has this weirdness, which I hadn't really thought about before, which is you only get to display one thing at, say, 7:03 p.m., right? 7:03 p.m., a bunch of smart people have to figure out, what are we going to display at 7:45 p.m.? And you only get one shot on what's going to be on that channel at 7:45 p.m. So there's a lot of thought into what goes on there in a way that when you're just putting out 300 URLs a day, you don't necessarily need to think about the freshness of that one time. So there's a whole bunch of interesting problems around making sure that people are making informed choices about what is going to be scheduled at a certain time, which I had never thought about before.
Mike Massimino: Yeah, and it can change. It happens to me a lot. Yeah, sometimes I'm down there in the green room and like, oh, we have to cut your segment because we're moving on from the space story because an announcement was just made here or there.
Chris Wiggins: So there's a whole bunch of levers that these people are pulling that I just never thought about. For example, who are we going to book? If something weird happens at 8:00 a.m., better call Mike.
Mike Massimino: Yeah, yeah.
Chris Wiggins: Mike is always going to talk to you. It just never occurred to me. There's a lot of levers that go into producing linear TV that I just never thought of.
Mike Massimino: Yeah, a lot of same-day requests. And that's where the remote hits have been helpful.
Chris Wiggins: Yes.
Mike Massimino: Where you know, hey, can you come on at three o'clock and it's 11 in the morning? You're like, yeah, okay, I can just do a Zoom. I can't get to the studio, but we'll do a Zoom. So you're helping with that as well? You're involved with those things?
Chris Wiggins: These are things that I'm starting to think about, meeting those people. And part of it is just like when I started thinking about biology, I had to meet biologists and ask them a bunch of very naive questions. And so that's part of the excitement is if you find yourself in some situation in your life where a bunch of people who are expert at a thing are willing to field your dumb questions about that thing, that's gold if you like learning.
So if you like learning new stuff, what a great opportunity it is to be in contact with a field that has expertise. You have none of that expertise and they're willing to sit through your really dumb questions about that thing. That's wonderful.
So yeah, it's been a whole education meeting this whole set of people who are producers for shows and sitting in and watching shows get cut, looking at what software they're using and how they're making decisions either in the calmness of 3 p.m. thinking about 7 p.m. or in the heat of, you know, something weird happens at 6 p.m. and then you think differently at 8 p.m. So that's been a whole education for me.
Mike Massimino: Yeah, when you get shown a different world, whether it's entertainment or whatever it is, it's really interesting to get in and see there's some commonalities too. I think a lot of it is the teamwork nature of it, the passion that people have for it. I think in successful organizations you see that there are some commonalities in those things.
Chris Wiggins: It's good to be a researcher. Being a researcher is an excuse for lifelong learning. You can always learn more stuff.
Mike Massimino: Is it fun being reunited with some of the folks from the Times that are over at CNN?
Chris Wiggins: Oh, it's a hoot. It is a hoot.
Mike Massimino: Yeah. You know, as I find myself getting older, you start, people that you worked with before show up again and you have those connections. It's always great.
Chris Wiggins: Yeah, it's absolutely true among my research colleagues too. Like, if I get to see somebody I went to graduate school with or I published a paper with 20 years ago and they're on a different continent or something, it's a hoot to see people and see, you know, if you're lucky enough to be in a career where you get to hang out with smart people who are thinking about similar things from a very different vantage point, right? It's very educational.
Mike Massimino: Yeah. You put yourself in a position with like-minded people with similar interests or passions and you'll see each other throughout your lives from time to time and reconnect. Yeah.
But what's your next moonshot, Chris? Is there something else you're thinking of down the road or what you're working on now that's the next thing you want to do?
Chris Wiggins: You know, I continue to put myself in challenging situations where I'm not entirely qualified, which is always a way to learn things. So, one is I've been working with the Dean of Columbia College trying to think about how can Columbia College think about AI in the curriculum? What should be the standards that faculty might have available to them in terms of tooling and expectations?
What should students expect about their experience as Columbia College students? What is, you know, it gets us an opportunity to think expansively about what do we want college to be? What do we want a graduate of Columbia College to be equipped for? So, that is definitely a challenge, a dynamic challenge. In research, I've been working quite a bit continually with computational biologists and experimental biologists to think about how do we learn more about the way basic biology regulates itself armed with abundant data? And in CNN, as we discussed, there's a lot of challenges to help a community think about can success be quantified? And if so, what are the right ways to drive success quantitatively using machine learning?
Mike Massimino: Chris, it's been great having you here. We're very lucky to have you at Columbia. Just by all the students, as a colleague, I've really enjoyed getting a chance to hear your story. And I think anyone who's listening can take a lot out of it. So, thank you very much for what you're doing here, for the world and for Columbia and for taking time out of your busy schedule to speak to us this morning. Thank you.
Chris Wiggins: Thank you, Mike, for making time.