E80: Financial Plumbing with Patrick McKenzie and Erik Torenberg

Hey, Upstream listeners. This week we're airing a conversation I had with Patrick McKenzie on Patrick's show, Complex Systems, on the Turpentine Network. Patrick, also known as Patio11 on the internet, is a polymath who writes online and is an advisor to Stripe in addition to hosting Complex Systems. If you haven't already, go check out Complex Systems for more like the episode ahead. Patrick and I discuss map versus territory problems in big tech, What the banking industry does, and money laundering. Please enjoy the episode.

Patrick, excited to turn the tables this time and interview you and get into some of the Bits About Money classics.

Thanks very much, Erik. For those of you who don't know, I write a newsletter called Bits About Money, usually about the intersection of finance and technology. It's at bitsaboutmoney.com, and we're going to talk through some of the things I've written about over the years.

Excellent.

So I want to get into your Seeing Like a Bank post, aptly named after Seeing Like a State. But before I want to do that, I really want to talk about this idea of Seeing Like a State talks about sort of the problems of centralized control and operation, but in big tech, seems to do very well. And so I'm curious, if you wrote a post or a book seeing like big tech, what would that be?

I think that is a book that is begging to be written. I'm not sure I'm the person who should write it. One, I'm very happy that the work and the mental model that is Seeing Like a State has kind of metastasised into the tech industry and to the intellectual communities around us. But I think there's a bit of a dialectic in that thesis. Seeing Like a State, if you haven't read this book, by the way, the author passed away recently. It is very worth a read. The brief version of the thesis is that high modernism tries to reduce the complexity of the world into a set of artifacts that can be operated on by government planners. And because reducing the complexity of something necessarily shaves off details at the edges, this has caused a bunch of human tragedies over the years as government planners have decided to do things that are either disrespectful of the culture of the people that they are serving, or in some cases, like German forestry in the 1700s or 1800s in the first few chapters. And the planners start to believe a different version of how trees operate than the way trees operate in the real world. And so they attempt to impose their forestry vision on this and successfully kill the forests without meaning to do that. So the and yet version of seeing like a state is, and yet we live in a society which is heavily indebted to the intellectual effort that is high modernism. And yet we do have these enormous bureaucracies around us. And yet the IRS, Social Security, Your water department, et cetera, observationally exist. And while you have problems with them, it is not the fact that everyone in your city has died because it is impossible to run complex systems. And therefore, the takeaway from seeing like a state should not be it is impossible to run large bureaucracies, but that there are dangers in running large bureaucracies and specific failure modes available to them. Does big tech prove for all time that It is possible to make bureaucracies that are much, much larger than previously extant bureaucracies and execute on them well? Yes and no. So I think observationally speaking, the App Amigo, FaceSoft, pick your favourite of the champions, they do some things that are strictly outside the capabilities of any other human institution, present or past. And so if you look at the supply chain that is Apple, it's a good thing Apple doesn't ship bullets around, because with the possible exception of the current US military. Apple probably has the most impressive supply chain assembled in human history, or something that is very close to it. And if you don't believe Apple is that, then Amazon is that. Pick your poison. They are just so many leagues better at shipping material from point A to point B than the US was in, I don't know, World War I, Civil War, et cetera. Pick some point at which the US was obviously a major nation and yet could not successfully arrange for a fleet worth of ships arriving from China at an arbitrary port on an arbitrary day to get on the airplanes to successfully make a launch day that was announced a year in advance. So big tech is hypercompetent with respect to some things, but big tech is also observationally not omnicompetent. We can look at things like Google and say, okay, Google's had a pretty good run of it. They've done some very good acquisitions over the years. There are global-scale products that they operate. that are by far the best with respect to those things that they do. And they also have a long run of failures. And it isn't obvious to me that Google would successfully execute on forestry management, where 18th century Germany was not able to successfully execute on forestry management. Is forestry management more something that rounds to YouTube, or is it more something that rounds to Wave or any of the other— pick products that Google has killed over the last 10 years. There's certainly enough examples to choose from. If you were to write, seeing like big tech, hopefully there would be some teasing out of why and what are the commonalities between the differentiated but they rhyme group of hypercompetencies that these companies bring to the table? And then what are the things they consistently fail to do? I think nobody at big tech thinks that they have cracked the final word on compensating and incentivising employees in such a way that it maximises for the interests of the organisation versus the interests of the employee. The principal agent problem, still a problem, still working on that one. For those of you who haven't heard this said many times by people in and around tech, one of the reasons it is so difficult to get new products off the ground, or rather one of the reasons it is so difficult to sustain new products after they've gotten off the ground, is that the compensation system that Google goes to war with is one that heavily incentivizes you putting your name on a new thing and heavily disincentivizes you being the caretaker of that new thing in year 4. And as a result, Uh, the A-team at Google either works on the, you know, Google Search, Gmail, et cetera, the things that have all the users in the world, or they work on whatever Google's, like, brand new messenger system is as of the time you're listening to this podcast. There will be one. There will always be one because Google knows institutionally, if we just made a competitive messenger, it would have literally billions of users and it would be literally one of the most valuable applications that was ever shipped. And Google knows that so much. that they have made 20 cracks at creating that product and all of them have failed. Does that necessarily mean that Google will perpetually fail on messaging? I don't know. But it is a very important data point that Google, despite having all the king's horses and all the king's men, continues to fail at doing that.

Yeah, the incentive problem is a big one, as you mentioned. Another problem people often bring up is the information problem. Going back to Hayek, this idea of it's hard to have the information of an entire economy, whether it's of a microeconomy or the actual economy, and it's the pricing mechanism that gives better information in a decentralized system. But because of AI's power to process information, maybe that will give more of an advantage to centralized systems that are trying to process and interpret Massive amounts of information that previously only a decentralised system could enable. Are you sympathetic to that?

It will be extremely interesting in seeing how that plays out. And I think that based on the current capabilities of AI, it seems very likely that the task, you know, run a for loop with a minimal amount of near-human attention attached to a very, very large amount of relatively similar things, like, that seems like it's going to get much easier than it was in the status quo. This is one of the things that states are historically really good at, like, right, Run effectively a for loop over a tax return times 300 million people. That's the problem that tech says, that's the thing that doesn't scale. And the government says, well, that's funny, we do it every year and it only takes us like 80,000 people. That's barely Google. So capability of AI is like run a for loop that goes over every human utterance that is entered into a computer system. Probably something that will be, one, a little bit terrifying, and two, terrifyingly within the reach of an arbitrarily well-resourced organisation within either the status quo or at most a couple of years from the status quo. Does that necessarily give you complete control over reality? Probably not. There's some Claude Shannon information theory-esque stuff going on here where if we are okay with truncating the dimensionality of the artefacts that are in front of us, then we can process a very large number of artefacts, but we will need to truncate. If we're not okay with truncating it, if we— genuinely care about artefact number 17,631 and the ways in which it could either present differently or the experience of looking at that artefact would need to change our model for how all artefacts are done in the world, then I don't think we have clear line of sight to the magic technology that absolves us from the responsibility of actually looking at it and seeing whether it should affect our mental models.

Relatedly, Peter Thiel famously said, that AI is communist and crypto is libertarian, or AI is centralizing, crypto is decentralizing, or AI gives power to incumbents and crypto to startups. And I think we've seen AI empower incumbents. I don't think we've seen crypto do much of anything yet except to enable speculation. Not to criticize crypto too much, but I haven't seen a libertarian or decentralizing renaissance thanks to crypto just yet.

Yeah. I sort of shrug with regards to like the political implications of various technologies. I think that's kind of heavily overblown. And the more interesting question to me is who is using this and what are they using it for? And, uh, like observationally, there is not that much uptake of crypto in the world relative to, uh, like tech systems that are truly in broad use. And AI, the shape of the graph for particularly the last, um, The earlier half of the last 3 years looks very exceptional to people who have been around tech for a while, like fastest growth curve we'd seen for a while. Plus, in addition to the uptake growth curve, the capabilities growth curve was somewhat mind-blowing. And so continues to be a thing that I'm very bullish on. I think many informed people would say that it is relative to the previous rate of surprise. The degree of surprise in the last 12 months has not been as surprising as the 24 months previous to it were, which is potentially a good data point if one thinks that one continues getting surprised right until the end of the world. But that is also something I somewhat shrug about.

We'll continue our interview in a moment after a word from our sponsors.

Yeah, let's transition into seeing like a bank. So you write in the beginning, banks are extremely good at tracking one kind of truth, ledgers. They're extremely bad at tracking certain other forms of truth for structural reasons. Why don't you unpack that? What are those other forms of truth and what are the structural reasons why they're bad at it?

Sure. So aspirationally speaking, and it is an aspiration more than a law of nature, banks are relatively good at recording transactions and relatively good at making sure those numbers sum at the end of the day and not losing all the money. These are the function of ledgers. We've had them since Renaissance Italy, the Medicis, double entry bookkeeping. One of the most underrated technologies in human history, by the way. It is why organizations work at scale. It's up there with writing. Be that as it may, I think many people have had the experience of if you have a relatively routine issue with your bank and you go into a branch and call them, you end up getting passed around different divisions. And every time you are passed to someone else, you have to re-explain what your issue is. You get told conflicting things in different places. And Maybe you get a satisfactory resolution, but it takes you hours where it feels like it should be something that can be done in minutes. Maybe you don't get a satisfactory resolution, and maybe you have some toothy consequences to you as a result of this not happening. That's the sort of thing that banks have structural issues for that to creep up in the user experience. Why are these structural issues happening? So if you were to peer behind the curtains of a bank, There might well be a customer service division, but it is only one unified customer service division for— in one tiny sense. And that is the nameplate on a building that there are multiple floors of people organized into multiple teams. But you drill into what it is actually doing in terms of organizing those people to deliver services to customers, it is typically multiple silos of people who are arranged in a tier structure. Where the least senior employees, who will often be outsourced, but pretend for the moment that this is all one building in Middle America, even though that is not the dominant model for most people's banks at this moment. The least senior employees are in tier 1. All phone calls go into tier 1 to start out to essentially triage as to whether this is one of the very easy calls that someone with the least amount of power in the bank, but the cheapest ability to keep on the phone, can quickly deal with, or whether it needs to get passed up the chain. And so those very easy calls are, you know, what's my balance, et cetera, et cetera. And you would think, what's my balance is entirely answerable by a computer in 2024. And you're right. But at the scale of the bank, stupendous number of phone calls are, what's my balance? Do you have the money that I thought I was going to get today? Et cetera, et cetera. Those sort of things. Real, a shocking portion of phone calls are things which you as a natural human being would expect the bank to never get asked, but they get asked all the time. And so given that the bank is quote unquote wasting its time those calls, they would prefer to waste the time of someone who makes like $12.50 an hour versus wasting the time of a senior executive. So tier 1 exists to protect the quote unquote expensive people in the bank from needing to talk to people who have interesting expectations about what the role of bank is in this relationship. Man, maybe I'm being a little too euphemistic about this. Think of the worst breakup story you've ever heard in your life. Then think of someone who is recounting that breakup story and needs to tell it to someone and then decides for whatever reason to tell it to the bank customer Like, happens every day by the thousands. Some people just need to talk. Some people just need to have someone to yell at over, you know, one would think about, well, they've chosen the bank to yell at because it's a money issue. One, they often yell at the bank for things that are not strictly a money issue. Two, they'll yell at things which in their mind are sort of like money issues, like often not having enough money. And the bank will be like, well, we are indeed the money people. And, you know, we have a flowchart for how much of your venting we have to listen to before we are allowed to disconnect this call. That, by the way, is a rule. But the bank has this iceberg of issues that you don't think exist because you haven't had to submerge yourself in this problem. Tier 1 protects the rest of the bank from that iceberg. Tier 2 is typically people who have served a little bit of time in Tier 1. Much of Tier 1 decides after a year or 2 of doing this very stressful job where they get yelled at all the time for things that are out of their locus of control, that I don't really want to work in a call center for the rest of my life, and they bounce out and do other things with their lives. Tier 2 has survived. They have decided, okay, I'm going to become increasingly an expert at a relatively narrow part of the bank's operations. So instead of answering any question about checking accounts, I'm now going to be the, you know, business account specialist or similar. You know, the exact breakdown depends on the financial institution. And so Tier 1 has prescreened the calls and then escalated some portion of those calls to Tier 2. And hopefully at the time it ends up at the business account specialist's desk, Tier 1 has verified, yes, this is indeed a business calling. And 2, it's not a crank. Some business owners are cranks. What are we going to do? And then Tier 2 has more options than Tier 1 does to affect resolutions. So for example, we talked earlier, like, aspirationally, the bank should not lose all the money all the time. Banks will lose money all the time. The magic word for it, if you are reading disclosures, is an operational loss. There are infinite sort of fact patterns in the world that will result in an operational loss. But from a user's perspective, the important thing is I expect banks to always get money from point A to point B when I want it to, and you didn't get the money from point A to point B. And sometimes what will happen in the bank is like, yep, that We don't know why, but money that went in at A and should have gotten to B already did not get to B. And the thing that the bank will often do in this circumstance is like, okay, charge it off to operational losses. We will make good on that money and put it directly in B for you. And they don't trust the folks that are making $12.50 an hour to make this call in some cases, but tier 2 might have an authorization available to them of if I'm the first person a business talks to that has authority to actually make an accommodation on their behalf. I might have $1,000 before I need to involve my boss, and maybe $10,000 before we need to get it out of this tier of the call center and send it to somebody else. And below that, it's just like I have a button, you know, type a tweet-length reason for the message, etc., etc., hit enter, and then money gets credited to their account. And that bank— that isn't the bank magically pulling out of thin air. They are, you know, Taking a loss in the service of their agreements to the user. But importantly, when a user gets passed from tier 1 to tier 2, many banks, they either haven't made the IT investment or haven't made the operational investment for tier 1 and tier 2 to be able to share any context between each other. And so from the user's perspective, I just spent 35 minutes on the phone with someone who felt like he doesn't know anything about anything. That's not strictly speaking true, but okay, it rounds to true. Like the human capital issues for tier 1 are what they are. And now I have to, like, repeat everything I've said, and the bank has forgotten everything in between these last 3 minutes. Structurally speaking, yeah, that's true. The bank no longer has active recall of anything that you have said over the last 35 minutes. You might get passed off to someone else. You might get passed off to a different department, et cetera, et cetera, at this point, and you will have to explain yourself over again and again. Why? Is it that banks hate you? No. Is it because they are trying to optimize for their profit margin? No, it's sort of an unsolved problem. We were talking about the difficulties of seeing, like, a state and the necessity of no map can represent the territory because the map by nature is smaller than the territory. For arbitrarily complicated territories, it is an unsolved problem in human affairs to successfully shrink that down to only the correct number of dimensions and then pass that off arbitrarily. The largest banks in the world that are best operated with regards to their customer service systems can do a tiny fraction of what is implied by this, and they can do that because they They spent billions of dollars on that problem specifically. If you're banking at, like, First National Bank of Kansas, they haven't spent billions of dollars on their CS team. You're going to get what you're going to get. Why are you okay with that? Well, think of your interactions with, say, the medical community. You have had the experience of working with arbitrarily competent professionals who've had many, many years to become specialists in a particular field of endeavour, and then they send you bills. And those bills sound like either you or someone on your behalf is going to have to pay $9,000 for resolving this problem. And there are relatively few problems with banks that you as an individual have had where you would be truly okay with the possibility of a surprise $9,000 bill just for the resolution of an issue, not the actual, like, money. And so what we do is this kind of, like, insurance mechanism where at a very low cost to users paid out of the net interest margin on their checking account or Out of the interchange that they have on their credit cards. We have an escalating tier system of people who are relatively cheap to talk to all the way up to people who are extremely expensive to talk to, and they will figure it out in expectation. And the experience of that is extremely frustrating to many users, but it is far cheaper and delivered at a price that people will countenance. And so this is the way the financial industry thinks it does. And so that is the seeing like a bank thesis in a nutshell with regards to And 10 years from now, do you expect this problem to be solved?

Or do you think this is just a repetitive— like, this is just how banks worked and just going to be continued how they work?

I think we'll continue seeing tweaks at the margin and some improvements here. It actually isn't the case that there has been no improvement in banking or retail financial services in the last couple of years. Like, the existence of mobile apps that give you an up-to-date account balance so that you don't have to call in and ask a human to read you your account balance out of the database Is a major innovation and improvement in banking services. And those will only get better over time, effectively only get better with some laggards, yada, yada. People don't perceive it this way, but the tiering system is actually an improvement over what we used to do, which is just your problem persists for forever, and hopefully someone at the branch can deal with it for you. But if they can't, well, banks, what are you going to do? Interestingly, if you look at old bits of fiction or old bits of culture that are preserved in amber, You will see occasional anachronisms like that where there was a bank issue and a widow lost her life savings, or there was a bank issue and, you know, Monopoly, bank error in your favor is a fun mechanic there. It doesn't happen much anymore because the rate of bank errors is like far lower. And of course, it would be a much less fun game mechanic, but one that very much happened in the 1910s. If, you know, bank error, the ledgers are wrong, you lost all your money, you're out of the game. Congratulations. Are we going to get better at operating the the tiering system. One, as human adjacent cognition becomes much cheaper than it is currently, replace some work that is done by someone who is relatively less than trained for like $12.50 an hour with— you can get arbitrary amounts of relatively less than trained or even potentially relatively trained human adjacent cognition these days. And after banks get good at deploying that, I think we might see efficiency in routing things from the first point of entry into the call tree, quote unquote, to some professional at the bank that can actually responsibly deal with that, that might get much more efficient. And so that will probably be good and people will probably still be a little unhappy with it because they had to redundantly type their stuff, one to a computer and then twice to a human, but they'll be talking to one human rather than 6 in a row trying to wend their way around the bank until someone correctly finds the desk that has a resolution issue. So I'm a bit optimistic about this, but evolutions rather than revolutions are typically the order of the day in financial services.

Yeah. You write about how society's goals conflict with banks being good at banking. Is there more you can say about that? And then maybe what can be done about it?

Yeah, there are some parts about this, and I don't know how to handicap this changing in a moment. One of the most frustrating customer experiences with the bank is when you are told, for reasons which we don't feel like disclosing to you, we're closing your account. This is not appealable. You and we should see other people, and we are deciding to see other people effective immediately, and we'll mail you a check in 30 days or similar. The exact punchline differs depending on how you arrived at the situation. What has often happened there is what is euphemistically called a risk-based decision, but the risk is ultimately generally not a credit risk-based decision. The bank is not worried about losing money on your account in particular. It's often a regulatory risk-based decision where they're, for better or worse, all regulated financial institutions are surveillance machines. They have to be by regulation. There's a department of people who are effectively working as intelligence officers within the bank. It's called compliance usually. And someone in compliance has, in a fashion assisted by computer program, but usually there's eventually a human making the call. And no, usually you cannot talk to them. They They've taken a look at your account and said, yep, that's a no for me. And it's often a no for them because they are responsible, seeing like a state, they're responsible for generating a large number of decisions over an infinite fact pattern that is the economy in the real world. And those infinite facts about a particular user are compressed into a small number of dimensions to show on their screen, and then they have to make a decision based on those dimensions. And they've decided to no. Why? Society has different goals for the financial system. Like, one of the social goals of the United States is we find ourselves in a geopolitical conflict with Russia at the moment. There are some people who are caught in the middle of that. And so one of the most frequently frustrating things that has happened in the last couple of years in the financial industry is, okay, the United States and Russia are geopolitical adversaries at the moment. Russia has done an incursion into Ukraine. The United States is not a geopolitical adversary of Ukraine. However, there is a territory in the Ukraine that Ukraine is not in constructive control of at the moment. There is a worry in the departments that are tasked with this in the United States that Russia could use its constructive control over those regions of the Ukraine to commit money laundering. And thus they tell financial institutions, don't service customers who are in these regions unless you do enhance due diligence on them, which the financial institutions hear as, if my institution does not have a huge specialization in doing business in the Ukraine, I should not be doing any business in the Ukraine because I will not be cognizant on a week-to-week basis of whether I am in compliance with that directive from the government or not. And we have a split-brain problem in the decision-making system that is the United States, where the government thinks it is saying, Please cut down on oligarch money laundering, which supports Russia's war effort. And what financial institutions are hearing is, "Got it. Definitely don't fund any scholarships to high schoolers in the Ukraine right now." That is a bit of an exaggeration. The things that I could say that were not exaggerations would you know be improper leaking of private financial details. But the hey, I understand what we're doing. I understand it's very negative. I understand there is an innocent person here who is consequenced by this decision we are making. But yeah, you know, my hands are tied. I have no locus of control. Our business will not dump the billion dollars that we would need to do business in the Ukraine in a compliant fashion. And so this person is going to have a negative experience today. Moving on to the next case. It is a thing that financial institutions have decided many, many hundreds of thousands of times over the last couple of years. Is that a decision that we want? No. Is that a decision that we are like priced into making again tomorrow? Yes, we will make that decision again tomorrow many, many times. And I don't know. There is no obvious patch on this issue, or it might have been patched in some places by now. There is no obvious situation that resolves the geopolitical conflict between Russia and the United States. There is no politically compatible with the economy that is the United States that says, okay, in society's ranked ordering of wants, we need this particular financial institution to exist to lower the cost of capital to farmers in Kansas. We also need it to be fair to Ukrainians. When those two conflict at the margin, I will optimize for being fair to Ukrainians based on not discriminating based on nationality or geophysical location in the world or similar, and pessimize for the availability of capital to Kansas farmers. No one is willing to say that in the United States of America. And so we make that— we have made the choice that we have made. And that's a little bit desultory, but a lot of the breakage that happens at banks has not really happened at the bank. It is the bank being made into a mouthpiece for a decision that has been made either in an identifiable elsewhere or in a less identifiable elsewhere that is just the United States existing as a system. A lot of the breakage that happens at the bank customer interface is simply describing this bluntly in a way that a bank will never verbalize to a customer. You are poor, and that means that You do not get what you want in life. I'm the person that has to tell you that at the end of the day if you make me tell you that. There are huge circumlocutions in the banking system to avoid saying this, but at the end of the day, you pay money for things in a capitalist system. If you don't have the money, you aren't able to pay for the things. The person that is most likely to tell you, bad news, chief, you don't have the money, is a bank officer at some point or someone connected to the financial system. That's one of the things the rest of the economy pays the financial system to do is to deliver that bad news so they don't have to, so that they sit— like a restaurant can say, your card was declined, which politely says it's the bank's problem, right? And you and the bank should talk through this versus you don't have money, and so I will not feed you and you will be hungry. And that kind of diffuse decision-making and diffuse responsibility for decision-making is one of the reasons the banks got identified as the bad guy. Which I think both happens on the retail level and on the social level. Like, they often get blamed for poor decision-making with regards to the economy generally, but that's neither here nor there.

We'll continue our interview in a moment after a word from our sponsors.

It's a good segue into another piece that you wrote about, which explains KYC and AML, which are a bit more subtle, a bit more nuanced than they might seem at the surface. Even though many people also just don't understand them on the surface. So can you explain what these mean and how they work?

Yep. So these acronyms, Know Your Customer and Anti-Money Laundering, are mandatory pieces of the international compliance regime since the early 1980s in the United States. And then this regime kind of percolated virally overseas, and it percolated virally largely on the back of the United States doing foreign policy using the dollar as a weapon, which is one of the things that I agree most with the crypto community on. Like when they make that complaint, they're basically accurate. I mean, you know, you can you can read the documents as these laws get passed as the not national but supranational governing bodies increasingly tighten the screws on banking havens, banking secrecy havens rather. Be that as it may, what is KYC? KYC is a complicated regulatory regime that says in substance regulated financial institutions need to know. Who controls accounts that do ongoing business with those regulated financial institutions, like who the individual person is behind it, some identifying information about them, and sometimes includes KYB, know your banking. What is the actual thing that you're transacting for? Sounds simple, deceptively complicated when you get down to the details. And KYC is one of connected to the reason why your bank account might get shut down if you happen to have a Ukrainian passport and that gets flagged as high risk at a particular institution. Or if I was I was once in line with somebody at a Japanese bank, and they were doing a wire transfer to someone they purported was their cousin in Africa. I have no reason to think that there was anything hinky about this at all. The bank has to ask, why are you doing a wire transfer to someone in Africa? He's my cousin. He's going to school. I'm sending him money so that he can go to school. And then the bank has to ask, okay, can you show me some documentary evidence that your cousin is in fact going to school? And the customer, for reasons, as someone who has supported family on many occasions, would often not have that documentary evidence at that point, right? And for unfortunate reasons, I'm not typically the kind of person that gets asked for it. Customer said to the Japanese banker, "I don't have a tuition statement." The banker said, "Oh well, sorry, we can't really do that." Then customer raised their voice. Bank dug in its heels. Yada yada. And you know, on that day. Someone went home unhappy and no tuition was paid, and that's downstream of their KYC and anti-money laundering considerations. The anti-money laundering considerations are, in brief, you will have been instructed by your regulators to come up with a list of low-risk, medium-risk, high-risk countries. Many countries in Africa are high-risk for money laundering, and to point at a particular one as an example without saying, like, this deserves to happen, Nigeria is on basically every high-risk list ever. And that's downstream of many reasons. Some of them are geopolitical in character. Some of them are bluntly reflective of the fact that there are very large organised fraud rings operating out of Nigeria that successfully siphon off hundreds of millions and billions of dollars out of innocent customers in other nations. And those nations would prefer that not happen. And they are willing to prioritise their potential fraud victims over the education of someone who is not living in their country. And no one ever makes exactly that comparison because the people that are coming up with the regulation do not have to make the concrete comparisons of which things to prioritise. They prioritise in generalities, and then the actual concrete consequences happen in a very different office many years later. And often there is a lack of feedback loop between those concrete consequences and the decision-making process.

Excellent. Well, I want to go deeper on Money laundering. Can you explain more about what's exactly happening when money laundering happens?

Sure. So money laundering is this very stretchy term, like taffy, intentionally. It basically means any action that you could take that would conceal from the state and the state's deputized surveillance apparatus in the financial system the linkage between the proceeds of crime and a transaction that is happening today. And the thing that you will get told if you do mandatory compliance training every year is that historically money laundering happens in 3 stages: placement, layering, and integration. Placement means in the traditional conception, crime usually has like cash proceeds associated with it. And placement is moving those cash proceeds into the regulated financial system in some fashion. And so if you can imagine like you're a 1980s drug dealer selling cocaine or whatever drugs they were doing in the 1980s, this is outside of my experience. You're selling that on the street corner. You have A big pile of dirty money. The money is not actually physically contaminated, although fun fact for drug dealers, it often actually is physically contaminated. But dirty money just means at the moment it is connected to an underlying criminal activity that we want to interdict, and there is no good explanation for this money that will pass muster with the financial system. So placement is, we need to figure out how to get that into the weakest link in the financial system, essentially. Then layering is, okay, if your money is in the financial system, but it's at the weakest link in the financial system. It's hard to do the things you want to do as a drug kingpin with your money. For layering, you're going to obscure the link between the weakest link in the financial system and the money that is now available to you or your designate or your entity or similar. And the traditional way to do that is, okay, maybe I deposited a debt at a particularly ill-regarded bank in a country that is known to be high risk, but I'm going to wire that between that bank and another bank and another bank and another bank and another bank and paperwork it as this is a distribution for intellectual property between 2 companies in the European Union, and then eventually send it as payment for trade goods between one country that is not on the high-risk list and an account in wherever I want to spend it. And the final step of the money laundering chain is integration. You now have money, but what can you do with money? You can't eat on it. Lying on it is kind of uncomfortable, especially now that it's electronic, and it doesn't accomplish the goals that you have in the physical world. So integration is, I want to take this money that I've obscured the paper trail about and turn it into something in the physical world. It will often mean buying assets that throw off clean money in perpetuity. And so, you know, notoriously a lot of Russian oligarch wealth has had the terminus of, and then it buys high-end property in either New York or London, and then high-end property in New York or London can be sold, and then the proceeds of that sale are definitionally clean. Where did that money come from? I sold the apartment that I owned in New York. Here's the paper trail. Oh, okay, well, you sold the apartment, or you rent it out to someone who works on Wall Street. And is their money clean? Yeah, it has an exequible source. It is legal to work on Wall Street. It is legal to pay rent to your landlord. Is your landlord allowed to charge you rent? Yes, we are a capitalist system. Okay, then that's clean money. And then you can use that to send your kid to private school in New York or buy jets or whatever you want to do ultimately as the rich owner of a criminal enterprise. So that is money laundering in a nutshell. AML regulations hit regular people because there is effectively no de minimis regime for money laundering, and there is definitely no de minimis regime for anti-terrorist financing, which is often conflated with anti-money laundering. The people who are in charge of this at various companies and in various offices of the government will say explicitly, no, we had a conversation about this and we have decided that if you send Literally $1. We are like, that $1 was a crime, and we will Hamas, maybe.

I don't know.

But we care about it from the first dollar. And so since we care about it from the first dollar, banks need to surveil transactions down to literally the smallest transaction that is representable on a computer. And that results in this enormous ongoing tax on the productivity and occasionally things like creeping out of the woodwork to hit people. And consequences on this, like, they're sometimes funny. Like, if you— notoriously for Venmo transactions, if you put in the memo field of a Venmo transaction, I'm spending this on cocaine and hookers, there exist places in the United States where one of those purchases is allowed, the other one definitely not. And someone in the vast surveillance operation that Venmo has to operate will have to flag that cocaine bit. And then, you know, a state machine happens. And at some places, they will tell you, okay, we understand that was probably a joke, but you need to stop making jokes like that. At many places, they will say, we're not comfortable with telling the customer that we think it was probably a joke, because that sounds like tipping off the stupidest drug dealers in existence. And there are many stupid drug dealers in the world, and they do use the financial system. And we will avoid creating a fact pattern that looks to government investigators like we were assisting drug dealers in avoiding the AML regime and instead cut that customer off immediately. And there have been financial institutions which perhaps, like, the soul of the institution was not on board with assisting the drug dealer in money laundering, but there was definitely, like, a captured subgroup of the institution which was very on board with that. And so that fact pattern in particular is something that institutions are extremely aware of. I believe it was HSBC that was fined $2 billion in the placement stage of laundering dirty money in Mexico. There were, like, a lot of bills that needed to be passed through a teller window, and there was a particular size of box that the cartel had standardized on because it's a large capitalist organization that wanted to have standards. And that box didn't fit through the teller windows, which caused a problem for capitalist organization A and their major supplier. And so their major supplier said, you are a big customer of ours. We will resize our teller windows specifically so that we can receive your cartel boxes of dirty money. Go figure. Government investigators, when they found out that this had happened, were pretty pissed. And thus a $2 billion fine. And thus, you know, every compliance officer for the rest of time being told about this $2 billion fine and told, like, one, these are responsibilities that our society has decided on for us. You must fulfill those responsibilities. 2, don't create ambiguity on whether you are assisting someone who is engaged in money laundering in their money laundering. That ambiguity will be read against us. And so locally, you know, locally explicable decision-making that often feels to the people who are impacted like, this is crazy. Like, why can you get a gun pointed at you just for making a bomb joke at an airport? Well, we've got a bit of a chip on our shoulder about bombs in airports specifically, but that certainly doesn't feel to the person who literally has a gun pointed at them in the moment.

Yeah.

You also mentioned in your piece another thing you agree with the cryptocurrency enthusiasts on, Which is this idea that the financial system has been deputized to act as law enforcement. Why don't you explain that and talk about the trade-offs?

Sure. So you could fill books with the requirements documents that AML and KYC regulations force you to do, but the short version is that one, the number of transactions that your institution will be doing that you are deputized into having sort of understanding of is far larger than you could hope to track just on paper. For similar reasons that the number of transactions that financial institutions take care of these days is far larger than one could hope to do if it was simply done with paper notes and smoke signals. We've got a computer program. There are a couple of vendors for the computer programs. Perhaps your institution has engineers on staff that are tweaking these around the edges, but largely this is coming down to a small number of relatively trusted vendors. And they are firing alerts based on signals, intentionally vague for a reason, that transactions are coming up. And when an alert happens on a transaction, you know, because you've been in the business of doing this for a bunch of years, these are not the choosiest computer programs in existence. Their false positive and false negative rates are estimable, and the false positive rate is like 99% plus, plus, plus. So you have floors upon floors of people at your financial institution whose job it is to get essentially tweet-length summaries of transactions. And then they need to decide, is this in the 99++% case, like, just a false positive and I let this transaction go through? Or is this potentially evidence of a crime? If it is potentially evidence of a crime, they, by regulation, must file this thing called a Suspicious Activity Report, which is essentially— you can round it to a 2 to 4 page interoffice memo, which goes to the Financial Crimes Enforcement Network. And when many people hear that, they think, ah, so we have this floors upon floors of people at the bank, and then we send it to FinCEN, that's the acronym for the Financial Crimes Enforcement Network, and then there must be floors upon floors of people at FinCEN who are doing the law enforcement on top of these memos. Actually not the case. FinCEN has a couple hundred people working at it. Most memos sent to FinCEN never get read, but they go into a database which is searchable in which many law enforcement organizations in the United States have the ability to search essentially at will. And those law enforcement organizations will tell you, it is hugely relevant to our operations that we have this large, large, large pile of memos available, not because we treat each one coming in as the same way that the New York Police Department would treat a— someone coming into a police office and saying, a crime was committed against me, but because they allow us to Reconstruct prior bad acts if we have an investigation ongoing. It's like the most clueful case for the utility of these memos. What this actually means is that there are tens of thousands of people, a population larger than the workforce of the entire intelligence community, who sit in every financial institution in the United States just clicking next, next, next, next, next, next, next all day long, and possibly, maybe or maybe not, At 3:30 saying, oh shoot, this one actually looks kind of hinky. I'm going to spend some time looking around our other records to put together a memo that probably no one will ever read. And so from a cost-benefits perspective, should our society be doing this? That feels like a close call to me. And there are downstream consequences of this, which I feel very certainly that our society did not countenance, did not sign up for, is not happy with. Like, if you think of the— if you forced people to articulate equity was all the rage the last couple of years before that bank banking the unbanked was all the rage. If you force people to articulate the concrete trade-off between banking the unbanked and KYC regulations, or between equity considerations and access to retail financial services and KYC considerations, I don't think people would draw the line where society has drawn the line right now. But that conversation has largely not happened. And there are many— pick what you value in the world if neither of those 2 considerations is very interesting to you. Like for me, economic growth, pretty important. I love when businesses get off the ground. And in those early days, I am acutely cognizant that they are often under-paperworked because paperwork does not spring out of the ether. Someone has to create it. The lack of paperwork causes businesses to have less access to financial services than they need to do the work in the first couple of days. And the first couple of days, weeks, months. And oh man, I've seen both in my own businesses and ones that I have professionally assisted over the years in various capacities that the The KYC hurdle bangs you very hard in the head early on. Far out of— you can come up with a reasonable model of if I help 100,000 businesses over the years, how many of them are actually a drug dealer front? And the answer is probably zero. But we exact a tax on the time of attention of 100,000 entrepreneurs to maybe potentially interdict a money launderer ex post facto and to generate seizures from them. That, by the way, extremely important bit of US practice. One thing that the KYC regime will tend to generate is it will tend to generate a self-contained crime on a particular document with a bank and make life very easy for the prosecutors. Lying to banks is a crime. You are forced to tell the bank, like, what is your purpose for using this account? You did not answer that question truthfully with, my purpose for this account is to launder my drug proceeds. And since you lied to the bank, we get to keep all the money that went through that account and file a forfeiture action against you. it. And it is easier to win those forfeiture actions than it actually is to convict someone of drug smuggling and similar, because you will be able to sort of prove the negative. The excuse they used was that there's a real estate deal. The real estate deal is as real as dragons. You can't actually point to the real estate deal in real life once it gets looked at by a prosecutor and the prosecutor says, okay, like tails, I win all the money and we will forfeit it from you and then buy the investigating law enforcement agency a whole lot of military surplus. It's one thing that forfeiture often enables at scale. That's the success case for these regulations. I understand how the people who benefit directly from that think that's a success case. It's very not obvious to me that that is a success case at the margin for society. And so I would probably draw the— we have a dial of what transactions that we expose this regime to, et cetera, et cetera. I would think that at margins that are relevant to me, we could be tolerant of more abuses of the financial system in return for getting an outsized increase in utility of all legitimate users of the financial system. And that, by the way, is one of the catchphrases which might have stolen from Dan Davies. But a thing that I've said in print a number of times is that the right amount of financial fraud is not zero. The right amount of money laundering is not zero. Like, that the financial system, to be usable by legitimate intent users, has to facilitate criminal activities. Very few people in Washington or in bank boardrooms are willing to say in as many words, Yes, if we operate a bank, we will facilitate criminal activity. That's good, because operating a bank is good. And yet this is something that we've priced ourselves into. And so the relevant question is, at the margin we are at right now, are we facilitating enough criminal activity? Which, oh man, if you didn't like the first sentence, you're really not going to like that sentence. But we are a society with— the system is complicated. Our desires are complicated. We shouldn't be monomaniacally focused on interdicting crime. We should also care about an immigrant in central Japan who wants to send their cousin to school. In the political economy of, say, the United States or Japan, that person has very little voice. But we as a society are in charge of, in some way, like allocating them some portion of society's attention to the question of, should you be able to use the banking system? I think, yes, they should be able to use the banking system for straightforwardly legal things. So that is my little rant, which rhymes with things that the crypto folks have been saying for a while. Things that does not rhyme with crypto folks is that simply circumventing the system will not fix the system, will not fix the outputs of the system. It will make a tremendous amount of money for some people who are circumventors, good faith circumventors and bad faith circumventors alike. And then eventually there will be a conversation which happens, which is like, okay, you guys have to go legit now and you can choose. Either you play ball with this regulatory apparatus, Or we send you all to prison for forever. And the, you know, onshore crypto economy in the United States has said, all right, we're going to play ball. And I would think that if one wants to devote a huge portion of one's effort over the, over a 10-year period to doing this, like rather than inventing circumvention technology, which you'll eventually have to walk back, you should just put your points into, you know, winning the societal debate that we are not at the correct margin on Helping out good faith users versus facilitating crime. Facilitating more crime.

Well, maybe as a closing question, if you could wave a wand and change anything related to what we've been talking about, in addition to things that you've already said, is there anything else you would, you would leave our listeners with in sort of, uh, the ideal Patrick McKenzie world?

Oh man, there's many magic wands. If one had a truly magic wand, making it possible for us to be more honest about the trade-offs we are pursuing, which is perceived to be incentive incompatible in the broader US political system and in our regulatory agencies and similar. But we talk out of both sides of our mouth on that because there are all sorts of other things that are higher on our own priority stacks where we are capable of appreciating nuance and of kicking a question off to experts who can keep 2 whole thoughts in their head at the same time, et cetera, et cetera. We should just decide that, quote unquote, if I had a magic wand, wave that magic wand in society and just decide that the trade-off of financial services versus crime interdiction is one of those things where we think we are capable of considering things on multiple margins and making decisions on those margins. So wave the magic wand, make America better at being America.

I think that's a great note to end this podcast on and very reminiscent of podcast theme in general. Patrick, thank you for letting me interview you and turn the tables.

Thanks very much, Erik.

Upstream with Erik Torenberg is a show from Turpentine, the podcast network behind Moment of Zen and Cognitive Revolution. If you liked the episode, please leave a review in the Apple Store.