AI Credit Bubbles: The New Shadow Banking?

That'll be the difference between people who are actively using AI and leveraging it to create self sustaining, viable outputs versus people where their prompts will be a couple chats back and forth. Kind of rinse and reuse that because it does half the job. How much free time do people want to use to basically do the hard work upfront is basically going to be the question it's the week of October 27th. Welcome to the Learning Corner, a Precursor Ventures podcast.

Welcome to the Learning Corner by Precursor. I'm Mia Farnam and I'm joined by our GP and Managing partner Charles Hudson, where we will be talking through the articles and Goodreads that caught our team's attention over the past week. Our first topic comes from Nick Potgalitzky, whose piece Beyond True and False Teaching Students to Interrogate AI Unreliability explores how students can develop a more nuanced approach to evaluating AI generated content. In his substack, Nick draws on the three pillars of literary theory created by James Phel, typically used to critique authors and literature, and suggest the same framework can be effectively applied to critique AI outputs. Nick explains that these frameworks offer a way forward moving beyond true or false to ask richer questions about how AI constructs and presents information. So the framework revolves around three axes of unreliability. The first is facts and events so what the narrator claims happened, acknowledging the potential for falsehoods or omissions, knowledge and perception the idea that a narrator might misunderstand, interpret, or overlook important nuances and complexities. And the third is ethics and values. So recognizing that narrators can often misjudge or undervalue aspects of an issue, while AI is a narrator in the traditional sense, Nick argues that teaching students to apply this kind of critical lens to AI outputs is essential, especially before accepting everything at face value. He concludes the piece saying, quote, students need to understand that they're co producing AI output through their prompts, making them not just consumers of AI unreliability, but participants in constructing it as well. So Charles, how effective do you think this framework is in helping students and then everyday people develop stronger critical thinking and analytical skills when engaging with AI?

I really like this article because I do think it gives you a framework for how to think about this. And I don't know. Yeah, I feel like it's a really good rigorous way to interrogate outputs and I fear that I still meet a lot of people who type something in, wait for the answer to come back, don't even give it a quick scan and just say well that's the answer, even if it has citations. And so I think in the same way that people started to learn, hey, some sources on the Internet are more trustworthy than others. And just because Google presents it as a search result doesn't mean that it's true. And there are people out here trying to actively influence what Google shows you in your search results. I think people are going to develop the same kind of skills with chatbots. I think the harder thing is it's one thing when you're delivered back a link and you have to click on the link and try to assess does this look real? When instead what you get back is a full on answer or a document or thousands of words of text. I don't know that most people have the patience, rigor or time to fully interrogate this. So I think a lot about what's the temptation. I think the temptation is you could know that AI outputs are unreliable. I really should fact check this thing, but I got a meeting in five minutes. Need to know the answer to this thing or I want to go in sounding smart. I'm going to run the risk that hopefully I won't say something that's an obvious hallucination or silly fact. But you know, I use AI all day, all the time, and sometimes it gives me an up and I'm like, well, that's not right. And I know it's not right because I know intuitively, like what the answer should look like. And I'll say, hey, that's not right. And oh, you know, you're right, that's wrong. Sorry about that. I don't know that we're at the sophistication level yet where everybody's asking the kind of questions they think is, but I think we're going to have to if we want people to be responsible consumers and users of AI outward. I think we have this weird dichotomy where employers are like, I want to hire people who are skilled with AI, but I also want them to go to college and learn critical thinking skills and not use AI. I'm like, you're actually putting a lot of pressure on, on educational institutions to teach people how to use this stuff responsibly. And then I'm going to go into a corporate environment where the standards they learned in school, rigor and responsibilities who may not apply either because the company has different values or everybody else who works at that company didn't have the same training. And you know, you're seeing these examples of like people, you know, lawyers citing cases that don't exist. Consulting firms publishing reports with. The one thing that gives me pause is it's not like everybody really got better at using Google. It's not like people started using Booleans and quotes. I think what happened is Google was like, people are bad at searching. So on the back end, rather than having people fix their queries on the front end, we will fix whether it's typos or whatever, we're going to do it on the back end and we're going to figure out what this person's actually trying to solve for. And I wonder if LLMs will have the path, which is it'd be nice to think that most people are going to become more critical thinkers and smarter and more willing to interrogate the results. The other option is that the LLMs could just decide people are lazy and if we want to maintain some semblance of trustworthiness, we have to figure out at the back end how to fix some of these things. I don't know. A few years ago I read this book called CO Intelligence by Ethanolik, which I read it and I kind of got it and I was like, oh, I don't work this way. A lot of it's like, think about AI as a coworker partner that's going to work alongside you. You guys are going to co create things. And now that's basically what I do all day. And that makes a lot more sense to me now.

The one thing I will also call out is that while we're training ourselves to be better and more critical uses of AI businesses on the back end who work D2C, they're also reconfiguring their entire website. Everything they do that interacts to work better on the rails of AI. So like there are companies and startups that are created to bas strip your entire website and like reconfigure it so that agents will crawl and pick up on things a lot quicker so that it surfaces it up to you on the foundational elements when you're using chat. So you're abandoning search. So SEO maybe doesn't mean the same thing anymore. So basically how do you reconfigure it? It's going to be two things at once, which will continue to be something we have to watch the muscle. I mean, I know you just touched on it, but I know you've been experimenting a lot building your own AI tools. Like how much deeper do you feel your understanding of AI and underlying models has gotten through that process versus just being using it at a higher level?

I Think the biggest aha for me has been the more structured you are and your thinking, the better the results you get. Systems. I've built a bunch of tools for myself where the original version of the thing I wanted to build. The first file I wrote, I was like, this is pretty detailed. I think this covers most of the use cases. And then I hand it to an LLM. It's like, well, what about these 10 edge cases that you don't explicitly address in this document? I'm like, oh, well, you're right, I didn't address those 10 things. Yeah, I do care about those. It's funny, if you want a machine to operate on top of this thing, you have to give it very specific instructions. And I think what I've learned is like, oh, I have a lot of mental shortcuts that I use that as I write them down and make them explicit, they don't make as much sense to me. And like, part of this has been building these tools has caused me to go back and revisit some of my core assumptions or look at gaps and say, well, why do you skip that step? I built this reference request generator skill and Claude that like, figured out what people should I talk to if I'm trying to get build the more complete picture of a founder. And there's some things I used to believe. Like, for example, in a world where many more people have raised venture capital and not all of those relationships are going well, does the fact that their previous investors don't love them mean as much today as they used to? And also the person who wrote the check might not be as experienced as an investor. I'm like, oh, maybe I could change the weight or just the framework I have. And that's something that I hadn't really thought until I started writing down all of the steps of how I do stuff. So I. I find that instead of becoming easier, it's made me be much more explicit. And being explicit, sometimes I'm like, well, that's dumb. Or I no longer believe that that's the best practice I want to follow. I don't know, I feel like it's been really powerful.

You can see basically the impact of when you change one of those levers, how much it does affect the output on the other end.

I mean, the other thing I'll say is I've never really had a reason to build a bunch of repos in GitHub, and now I have a bunch of them and the ability to see what happens to the quality of your output as you Iterate on something over time with versions. It's been astounding to me. When I look at the first version of stuff I start with, I look at how it's like why did I even think the first version would produce a remotely interesting result compared to the thing I have now.

And that'll be the difference between people who are actively using AI and leveraging it to create like really self sustaining long term viable outputs versus I think people who are. Their prompts will be maybe like a couple of chats back and forth and they'll just kind of rinse and reuse that over and over again because it does pass the job and to your exact point in the beginning. I don't know like how much for free time do people want to use to basically do the hard work up front? Is basically going to be the question.

Yeah, I've like given myself a development budget of like. I won't say it's more time that I. It's more time than I anticipated. If I didn't have to work I would probably do it five or six hours a day and that that's not the state given my job and my family and and many other things in life that I care about. But I found that it's been a really interesting educational experience for me.

Well, our second piece is from the substack but this time different. Titled Is AI the new Shadow bank and then in parentheses. Yes, the piece references Zoltan Fozar's seminal work on shadow banking which revealed that much modern credit creation happens off official bank balance sheets in these opaque shadow markets. In this system, collateral, not cash, becomes the true form of money with assets like treasury bills or mortgage backed securities serving as the basis of for future lending and liquidity. The piece claims that a similar shadow credit system has now emerged around AI. But instead of mortgages and traditional financial assets, the collateral is AI hardware, coherent contracts and access models. And through this circular self reinforcing system, the same quote unquote stack of GPUs is borrowed against, resold, repledged multiple times just as mortgaged assets were rehypothecated back before 2008. The piece closes with a quote. The real miracle isn't the AI, it's that we've taught credit to dream again. Obviously wanted your thoughts on this piece?

Oh, I got this from my friend, we're in a GP group chat and he sent it and my first I was like this is giving me red string conspiracy theory vibes the first time I read it and then as I Reread it, I was like, this is tying together so many of the things I've been struggling to reconcile, which is it's not just the circular, hey, I'm gonna, I'm gonna put $30 billion into your company and you're gonna buy $30 billion worth of stuff for me. All these circular transactions. It's also kind of like what are all these underlying assets and like how are they being financialized for lack of a better word. The thing that I've been most worried about, which is there's so much of this stuff that's intertwined and interlinked. Do we actually understand how it all works? Like we have companies that are signing up for capital expenditures that there's no, as far as I can tell, no feasible way for them to finance them under any reasonable set of expectations. But they're making them in satisfy. And what does that mean? And what is the value of a collateralized GPU that's two generations old? And I do think that I said this to my friend a while ago and this is something that's really been on my mind. It is very hard to blow up the economy with equity. You can do it. We did it in 2000. It is very easy to blow up the economy with debt. And we did that in 08. And like if you ask me what's scarier 01 or 080808 way scarier. When equity markets seize up, companies have problems but the rest of the world kind of functions. When debt markets, it's like turning off the oxygen that keeps the economy running because even businesses that have nothing to do with AI will find, hey, I can't borrow money to finance this tractor for my farm or to get a mortgage. And, and I worry that like I don't know who's watching this. I don't know who's like watching the store. Last time we learned people we thought were watching the store, were watching the part of the store they could see that regularly, the banking industry, but didn't really understand the size. And I'm not going to say that like the AI related investments have reached the scale of the mortgage backed securities crisis. But I don't know, anybody looking at this stuff is saying something here feels off.

And I mean I love reading through I guess what quotes, our kind of walkthrough from like a shadow banking concept. I mean the whole premise is based on the fact that shadow bank is whatever the financial system treats as safe, liquid and flexible. And so I was starting to think of like at what Point did we all decide that all of the infrastructure of which AI is built upon is safe liquid?

I think it was because that stuff was so scarce and it felt like, oh, like GPUs capacity model access were scarce. So if you had access to it, of course it's valuable, of course. And it's valuable today. And it was hard to conceive of a world where those things would become so plentiful that they would decline in value. I hope we don't end up in that world. But I have like so many questions because right now a lot of this is being propped up by fundamental optimism. People are like, this is. This is all going to work. And what happens if even a modicum of doubt gets creeped into the system?

Well, that's my last question. Kind of like, what does that doubt look like and what form does it take? Because we've been talking about this infrastructure debt issue for months and I think now I'm starting to see more posts or out about it. But the real question remains is has it range investor sentiment and also does it matter and actions as it pertains to the broader industry overall?

I read like a similar doomsday article about how basically there's a bunch of data centers being built where the math simply like cannot work. And I think to me that's the thing I'm looking for is like right now data centers have been a consensus good investment because of the demands for power, Pew capacity. How could you bet against data centers? I wouldn't want to necessarily take the other side of that trade. But I think if you were to see visitors not pencil as investments anymore, either because they're built and not utilized, or they're built but people can't operate them profitably without more data center and GPU financing, the whole thing glides to a whole and those are largely, largely the debt market that's financing those projects purchases. Look, if we continue to come up with hundreds of billions of dollars a year, we can keep the party going. I'm just not sure that can happen.

Yeah, right now in the middle, it feels like we talk about this internally and I feel like we have a pretty grounded sense of like optimism in the questions that were made. It doesn't feel like my peers around me share that same question all the time. So I think there's so much more money to be spent here. Like, I don't assume that there's going to be a hold anytime soon. And also on the venture side, I think that'll be delayed as well. Well, we might have a hold on the credit and then venture will follow later. It's not one before the other.

When it happens, bad sure will do what we always do, which is we will overreact. When it gets bad, everyone will be like, stop investing. Turn off the tabs. That's what we do.

Well, we're all out of time today. Thank you, Charles, for stopping through. Thank you, our lizard, for coming in as well. Please like and subscribe and we will talk to you all next week. Bye.