Eric Bradlow, Vice Dean of AI & Analytics at Wharton, discusses the launch of the new Wharton AI & Analytics Initiative and how generative AI can be leveraged in education, business, and society to enhance our lives.


Investing in the Future of AI

Angie Basiouny: The Wharton School is leveling up its teaching and research in artificial intelligence and data science. The school has launched the Wharton AI & Analytics Initiative. It’s an interdisciplinary, collaborative effort to shape how generative AI will be used for business innovation and beyond.

With me is Eric Bradlow, a longtime faculty member in our marketing department. He’s an applied statistician, and many of you know him from his “Moneyball” podcast on Sirius XM.

Eric, you’ve been named as the vice dean for this initiative. Give us a quick overview of the initiative and why the school is launching it now.

Eric Bradlow: As our Dean Erika James says, it’s really one of those meet-the-moment opportunities for the Wharton School. We’ve been in the analytics business, if you’d like, for 20 years. A colleague of mine, Pete Fader, and I started the first-ever data science center at any business school in the world 20 years ago. We always believed that data science, basically using empiricism, using high-powered statistical models, was going to be a big part of the future of business because of technology and new data streams that emerge.

But what has happened, of course, in the last almost two years now is generative AI. Fortunately at Wharton, we’ve had a center called AI at Wharton for almost seven years now. We knew this future was coming. We didn’t know how fast and how ubiquitous it was going to be, but the goals of our center are actually quite simple. There are four major goals. First, we want to be thought of as the No. 1 knowledge creators of business schools in the world.

No. 2, we want to impact students, broadly defined. Not just students at the Wharton School or even the University of Pennsylvania, but we want to democratize AI and business education throughout the world.

Third, we want to have an impact on the C-suite. We’re a business school. What I would say is we’re training people to be the leaders in AI and data science, but not data scientists. We’re training people to lead companies through AI and data science.

And last but not least, I’ve always called it analytics for good — how about we do some good in the world through data science? Those are our four populations of interest: researchers, students, companies, and the community. And now is the time to have that impact.

Basiouny: Let’s talk about a couple of those buckets. You mentioned students. With this initiative, you’ve got this really unique collaboration with OpenAI, the makers of ChatGPT. Starting in the fall, our MBA students are going to have these creative enterprise licenses with ChatGPT. What are they going to create with them? What’s the end result?

Bradlow: There will be a time when Wharton will be able to present enterprise licenses to the entire [Wharton] population…. There are some technicalities there, but we’re working through them. The answer is you learn AI by doing it. That’s the only way to get people to do it. For now, every MBA, every faculty member, every Ph.D. student will have access to the newest versions of ChatGPT right on their desktop or phone. They can actually experiment with it, see what it can do to enhance their education. We don’t view it as replacing us as faculty. We view it as enhancing their education.

At the end of the day, this is what is going to be expected of all of our students when they go into the workforce. They’re going to be expected to know, “How do I create video using artificial intelligence? How do I do the proper prompt engineering so that I can get the right answers out of these large language model engines like ChatGPT?” The only way to do it is to experience it, and that’s what these licenses are going to provide. The good news about the enterprise version is the data is secure. What I mean by that is if professor Bradlow, as myself, uploads his course materials and creates a localized version of ChatGPT, it’s not being sucked into the corpus that they’re using to train the model, so people around the world do not have access to my information.

And let’s be honest, that’s the business model, because ChatGPT for most people is free. If you get the enterprise, it’s free. It’s like Google. Google search is free to you and me, but it’s not free to companies that want to advertise on it. ChatGPT is free to you and me, but it’s not free to Amazon, that wants to create a localized version of it. That’s the way it’s going to work. We at Wharton have invested money and resources in an enterprise license, so we can educate our students on how to be leaders through using artificial intelligence.

Promising AI Research

Basiouny: There is so much research coming out of Wharton about AI. Can you touch on some of the areas that have already been impacted by our research, and what you find to be the most promising?

Bradlow: I think research comes in a few different buckets. One of the buckets is, even though we’re a business school, there are people at Wharton who are statisticians and computer scientists like myself who do work on the mathematical models that underlie artificial intelligence. And just so all of our listeners know, AI has been around for about 60 years.

As a Ph.D. student, I learned about artificial intelligence. Now, large language models, which is taking a big corpus of information, trading them in real time at scale, and providing generative AI, that’s new. But the problem of predicting the next word or token that’s going to be used, that’s not a new problem. There are academics who are, in my narrow slice, who are what we call methodologists who will work on the methods. The much larger bucket for researchers is people who are going to apply AI.

For example, how do you use patient records? [Say] I’m in the health care management department here at the Wharton School. I partner with Penn Medicine. I see doctors writing handwritten notes on patient records. Well, those used to have to be human coded. Now they don’t.

I’m in the management department, I want to know how AI is going to affect teamwork. If I can just ask a chat agent and not my colleague, what does that do to camaraderie and building relationships with your team? There are researchers, many of us, who also work on applications of artificial intelligence. Then, of course, there’s the student piece, which is how do I get my job as a student done using artificial intelligence?

I view it as almost like a three-pronged laboratory, which is the methods part, the applications part, and the how-do-I-get-it-done part with artificial intelligence.

Basiouny: Our professors are always trying to pursue relevant research, research that has a business case. Can you tell me what the companies are asking us for right now in terms of AI research?

Bradlow: Yes, there’s a lot of it. Let me just give you a couple of specific use cases. Obviously one big area is in customer service. For example, can AI be used for chat bots? And what are peoples’ reactions to the use of AI-generated chat bots? Can AI be used to coordinate or collect audio and video recordings, make them searchable, indexable so that we don’t need humans taking notes. The AI engine can do it.

The third part is what most companies are interested in. They have data all over the company. I don’t mean just numeric data, which is historically how we thought about it. But I have physical reports. I have online reports. I’ve got Excel spreadsheets. I’ve got PowerPoint decks. I’ve got a SQL database. I’ve got a big data lake sitting somewhere. How do I bring all of that information together so that Angie, sitting at her desk, doesn’t need to be a programmer to query this database? She can just use the English language, or even a foreign language, to query this database and say, “What do we know about the following topic? If we were to make an investment in X, what does our data say about this?” You don’t know Python, but you want to do some analysis. Well, you can now do that analysis.

Are there jobs that will be replaced? Absolutely. That’s not, to me, the large part. The large part is how does each person use AI in their job? As my colleague Stefano Puntoni says, and my colleague Ethan Mollick says, “It’s AI and the workforce, not AI or the workforce.” That’s what I see as the biggest, most exciting use case.

Thinking Optimistically About AI

Basiouny: That’s something I think is really important to touch on, too, because AI has been described as everything from society’s salvation to its doom. What gives you optimism about AI, and what do you worry about?

Bradlow: Given we have a political election here in the U.S. in 2024, my worries are more on a personal level, which is around disinformation, and creating fake content, which is almost indistinguishable, if not totally indistinguishable, from real content. I was at a conference in San Francisco, and someone from one of the large security companies said, “Imagine you’ve created a million fake articles and load them into ChatGPT. Well, it’s going to train on those fake articles, and then to anybody using ChatGPT, it’s going to seem like the truth.”

It’s the disinformation, and the fact is sometimes it just makes stuff up. It will give you an answer. That concerns me in a society that’s got some polarization in it and a society where we’re relying on these engines to provide us answers. My view is that “relying” is the wrong word. We shouldn’t be relying on them. They should be decision support tools. But I can promise you, I’m not betting my company’s future or my academic future or my teaching future or my research future on what an AI engine says, because I know, I test it all the time. Some of the stuff it says is just wrong. It seems right, but it’s actually wrong. So, my concerns are around disinformation.

Look, I love action movies. I’ve seen every Schwarzenegger Terminator movie. I’m not worried about AI, cyber, whatever they call it, taking over all of that. I’m not worried about the young people today because they always get trained in the best of technology, and we have to democratize that. We need to do a better job of making access available. I’m worried about the 50+-year-old — I’m pointing at myself — who needs the training because of what AI can do. Where do they go? I’m worried that the people whose jobs will be displaced, many of them don’t have the training or won’t receive the training to make AI an asset for them. It will be seen as taking over their job.

Basiouny: What gives you optimism about AI for the future, and about this initiative?

Bradlow: Well, I hope this is what everybody does. My colleague Ethan Mollick says this the best: “Just go on and start using it.” Think about all the things you do. I’ll give you an example. Our interim president, Larry Jameson, asked me to speak to a group of alumni on the role of AI in education. I used to use TAs (teaching assistants) in the classroom to sit there and record class participation. Well, I don’t need that. I have audio data. I need people to help me do grading. It’s not obvious I need that anymore. Every year, “Oh my God, I’ve got to create an exam.” No, I don’t. ChatGPT can create an exam for me. “Oh, I need to create a simulation for the class.” Well, actually the generative AI lab that Ethan and his wife, Lilach Mollick, are starting can generate a simulation for me.

The way I view it is Wharton should be thrilled. Eric Bradlow gets to spend more time on research. He gets to spend more time teaching in a pedagogically interesting way. He gets to spend more time mentoring our students. His time can be alleviated because he can get the generative AI engine to do [tasks] he doesn’t have to spend his time on anymore.

This is my fatherly advice to everybody here: Think about the tasks that you do. Before you do the tasks in the humanlike manner you’ve been doing them until now, try doing them in a generative AI engine, a large language model engine. See the results you get, and step by step, you’ll notice in yourself, like I did, replacing the things that used to take me a ton of time but weren’t what I call the “heavy, brain, human CPU stuff” I’d like to spend my time on. That’s what gets me excited.

Basiouny: Yes, that’s something I have heard our professors repeat quite often, which is, “Think of it as a tool that can free you up from the banal, so that you can do the more exciting aspects of the job that you love.” That sounds good to me.

Bradlow: Well, that sounds good to me, too. The Wharton AI & Analytics Initiative, we’re extraordinarily ambitious. We want to impact the world. We want to impact society, businesses, researchers, students. And we’re going to do it. Then the real magic to it is through partnerships. So we partner with schools to deliver education. We partner with nonprofits, to teach nonprofits how to use them. Obviously, we partner with scholars around the world. We partner with companies and the C-suite. This is the time where the ivory tower better not be the ivory tower because we desperately need the application and the corporate partners to actually bring what we’re doing to life. I’ve never been more excited, and I’m honored that our dean asked me to lead this initiative.