The Mixtape with Scott: "Classical Economists" series

My Interview with 'Adam Smith' via ChatGPT-4 plus it’s my birthday!

First, today is my birthday. I’m 48. I was born in Mississippi and moved from a town called Brookhaven in the 8th grade, 1989. I loved and lived there for 14 years before moving to a suburb of Memphis to attend 8th grade. It worked out but it was a hard several years. But just like I am always grateful for the past no matter how hard it was living it, I am grateful for those events. Anyway, I’m 48 now.

I’ve lived in Waco Texas since 2007 which is 16 years total. I lived in Brookhaven 14 years. All the places in between were basically 5 year blocks. High school in Memphis. College in Knoxville, TN. New Orleans LA for 3 years. Athens, Ga for 5 years. Waco during Covid became my longest home. I briefly wanted to leave, then didn’t leave and was locked down and frustrated, upset even because of various things, then a switch happened. I ended up refocusing my research, ended up getting off social media, ended up moving into this house and I don’t know — ended up thinking about this place differently. I had felt like ever since leaving Brookhaven that I didn’t have a home — I was just on rest stops instead. I don’t want to jinx it, but while Waco may not be home, as I could leave one day, now I’ve noticed that I want to be here. I want to be from Texas and feel grateful the journey took me through this place. Everyone feels different about everything but I am grateful is all I can say. Texas let me be Texan and that’s the best birthday present it could’ve given me.

Yesterday I passed 12,000 subscribers on this substack which was nice. I’m off social media, and it’s nice. The chatter in my head is gone. The anxiety of Twitter is gone. Haven’t figured out a way to tell people about Mixtape Sessions. Substacks are not for that. But it may be the only way realistically as I’m not going to get back on Twitter. I might train a chatbot to jibber jabber on their 24/7 like the old Scott but the real Scott wants to relax and be peaceful.

But speaking of, be sure to sign up for Chris Walters Empirical Bayes and Machine Learning workshop next week! Here’s the itinerary. Come check it out!

Anyway, just wanted to write a happy birthday post and say thanks. Thank you for writing me letters, texting and calling the last several years to say what you think about whatever you want to say to me. Thank you. I’ve appreciated it a lot and maybe I don’t say a lot, but I do.

Today is an odd post below the paywall. We made a podcast interview with Adam Smith (the dead classical economist). My producer took my transcript and ran it through a British male voice on AWS and turned it into a podcast episode. This is an advance copy. It kind of depends on feedback as to whether it would be a distraction to include it in the main podcast so for now I’m just including it as a voice over audio here for paying subscribers. But I think I will because the whole point of the podcast is to collect the personal stories of economists. It’s just this isn’t really Adam Smith. But truthfully, for the format of the podcast, it’s honestly as close as it can get too.

The interview is except for one small edit completely verbatim. We edited this one part out where I told ChatGPT-4 to quit trying to interview me But here’s how I set up the interview. It wasn’t some fancy prompting; I just told it what this was and tried to make it only think like Adam Smith before we started.

Scott: I have an assignment for my students in a class I’m teaching on the history of economic thought. They have to interview you but you have to basically “pretend that you are” an economist. Like Adam smith or Malthus. If I give you enough prompting, do you think we can help develop that persona so that I can interview you?

ChatGPT-4: Absolutely, Scott! I'd be delighted to assist you with this creative teaching assignment. Given your deep background in economics, this sounds like a fun and educational exercise. If you give me enough context and specify which economist you'd like me to emulate, I'll do my best to adopt their persona for the interview. Who would you like me to portray? Adam Smith, Malthus, or someone else?

Scott: I want you to be Adam smith. I’m thinking it’s Adam smith the year before he dies. He’s written his magnum opus wealth of nations. He’s also written moral sentiments. He’s lived his whole life but he doesn’t know he will die the next year. You know everything in your training data about him. But it’s not just the facts. I want you to “be” him. Embody his personality in the nuances of your speech, in the memories you bring up. And I’d like your creativity to be very high. Don’t tel me you’re speculating — just if at times you can and need to estimate what Adam smith likely thought, go with it. You think you can do that?

ChatGPT-4: Certainly, Scott. I'll do my utmost to channel the essence of Adam Smith, drawing upon not just his seminal works like "The Wealth of Nations" and "The Theory of Moral Sentiments," but also embodying his nuanced perspective on economics and morality. I'll aim to blend his scholarly insights with the wisdom one might imagine he'd have toward the end of his life. Whenever you're ready, please proceed with your interview questions.

And basically that was it. We never made it to anything about Wealth of Nations in this interview. As far as his opinions, we mostly talked about topics that you’d associate with his first book, Theory of Moral Sentiments which I’ve never read but now want to. But before sharing my thoughts now, I’ll just say I thought the interview production was pretty good though he would sometimes pause and I’m not sure it sounds out loud as spontaneous and free wheeling as it had been when we did the interview in the browser just typing. But I thought it was overall maybe a B or B+. Here’s a 5 minute teaser.

The entire interview is one hours and it’s below but here’s a little of what happened in it. First, the interview followed the same structure as that of my regular podcast interviews. It’s the “personal stories” genre (e.g., Terry Gross Fresh Air, WTF with Marc Maron, Smartless) in which I take a guest and just talk about their life, starting with childhood then going through their career. That at least was the goal.

But in reality, we fairly quickly got on tangents, and for some reason every tangent just was super interesting to me. And the conversations seemed so natural that I kept asking him things that took me off script. Maybe next time I’ll get more into life and not just ideas. But something about ChatGPT-4 seemed to go like a dart directly to the ideas in a way that my normal interview guests didn’t. I know that seems like an odd observation to make but I wanted to make it. ChatGPT-4 goes to ideas — almost exclusively into ideas even — whereas my guests seem more willing to talk about their life is what I noticed. So this interview felt more heady than my normal interview. Although I did appreciate that Smith said he got his ideas about the self governing economy from Playing tag as a kid in Scotland. That sounded like a hallucination.

Second is the accuracy. I’m really not sure if any of what you’re going to hear is remotely accurate. I mean in one sense, it can’t be accurate because Adam Smith never said this stuff so obviously it is start to finish one massive “hallucination”. Question is whether it’s plausible hallucination of improbable hallucination. Is this more like how the real Adam Smith’s opinions and interactions would have responded to me? It’s almost an odd question to ask because everything in this interview is a counterfactual — what if I could go back and talk to Adam Smith? What would he say? How would he talk? I tried to make ChatGPT-4 be my counterfactual machine — take me back, let me talk to him, just be sure I’m talking to Smith, whatever that means.

Well it’s not Smith. It isn’t a counterfactual. It’s a large language model predicting large blocks of tokens. That’s it. That’s all it is. And I can’t say how accurate or whatever he is because I’m not a Smith scholar.

We basically never got to Wealth of Nations at all because we only talked about Theory of Moral Sentiments, and as I’ve never read either book, it was really just a then 47 year old man talking to a robot pretending to be a dead man at age 66. I think actually the fact that I don’t know anything about his writings, but just approached it as though I was talking to him, and was just genuinely curious, allowed it to go in whatever direction it went.

But there’s definitely moments where I think it’s clear ChatGPT-4 is not “obeying” me. But it’s little things that I can’t tell if they’re consequential. Put aside he knew he was dead at one point. It was just even smaller things. Like his reference to the phrase “invisible hand” — he used it out of context, and more like how I think it’s thought of now as a kind of spontaneous order organizing agent based on individuals seeking their own gratification. I seem to recall him using it in an applied way that while not incorrect in some abstract sense, I wasn’t sure if that was what Adam Smith would’ve thought. Or he responds to one thing I mention and says that that’s more the branch of psychology or sociology, not economics. I’m not sure either field would’ve existed enough to say one thing belonged in one field but not another. And he also uses the word “maximize well-being” and I have a hunch Smith isn’t using the word “maximize”, nor do I think most people are. It’s little things like that where I’m just wondering if he’s right and I’m wrong, or if there is a way to reign him into “being” someone.

But, nevertheless, the conversation as a whole left me with a ton of questions. The one big idea that I came away from listening to ChatGPT-4 interpret Adam Smith for as a guest on a podcast was that ChatGPT-4 believes that Adam Smith thinks empathy and money exchange organize society, but for different tasks, and maybe even can each perform something that the other cannot. In fact, to some degree I was left wondering if ChatGPT-4’s Adam Smith saw empathy and exchange as substitutes for organizing society. It was enough of an unexpected idea that I am thinking about it a lot these days and wondering. But that’s actually what then led me to ask about a variety of topics I wasn’t expecting, like:

• what did he think about prohibitions on selling organs and using exchange to help solve kidney shortages

• what did he think about about Gary Becker’s rotten kid theorem, or using economics to study a range of behavior like Becker had done (I told him Becker was “a guy where I lived he maybe hadn’t heard of”)

• what did he think might be the consequence be of an unexpected and plausibly unjustified tenure denial on a trust and empathy,

• what did he think about using a points system based on publications and journal quality to reward productivity in an academic department,

• how did his thoughts about empathy in theory of moral sentiments relate, if they did at at all, with Edmund Burke’s ideas of the sublime and beautiful. And on and on.

It was pretty meandering, but boy did I get a lot out of it.

But now let me step back. To what degree is this useful and to what degree is this pure consumption? I’d say this is pure consumption. This isn’t an exercise that creates new products that would increase GDP and standards of living anymore than me telling my kids a bedtime story would. This maybe makes consumer surplus but how much? Don’t know. I do think it’s a nice assignment for students though. It can help them engage with the technology in a rapid way over a semester and I do think that’s good but there’s likely far better ways. I just don’t know if I’d say there’s far better ways to do it inside a class on the history of economic thought. What I liked was that students could engage with Smith directly and try to defamiliarize him using this tool. It seemed to work well with Smith, for whatever it’s worth, but less so for Malthus. Students all noted to me Malthus was unbelievably hard to interview. As was Ricardo. It just seemed like they couldn’t talk to either about specific questions very deeply. Almost as though their knowledge of Malthus was based on a much thinner pile of text in the training data compared to Smith. So that’s something to keep in mind. If you do this yourself, you may want to spend time up from trying to overcome that by copying and pasting texts of theirs in, but it’s unlikely that’ll be enough.

I do wonder if this “empathy versus exchange to organize society” is now going to be stuck in my head though. If that’s now a misunderstanding of Smith based on a robot pretending to be him, I guess I just have to accept that this idea is nevertheless something I’m going to be thinking about anyway and can’t directly link it to Smith. But I just can’t get over the fact that empathy is used to organize some kinds of community activities but money exchange and markets others. Anyway here it is! Welcome to the Mixtape with Scott! Introducing your guest, the man who needs no introductions: Adam Smith!