Optimization, AI and Opting Out with Coco Krumme
34:23 Share

Walking the money for the rest of us. This is a personal financial on money, how IT works, how to invest IT and how to live without worrying about IT. I'm your host, David stein. Today is episode five or five is titled optimization A I and opting out with coco chrome.

In this, our final episode of the year, I have a conversation with coco chrome, whose work I reference this past may in upset for eighty and optimization in that episode I mentioned mirvan king, former governor of the bank of england, wrote, as humans, we don't optimize. We cope yet. There's a great deal of optimization occurring that is brought incredible benefits to humans.

There also costs. Coca chrome describes ed herself as a writer, applied mathematician, an occasional filmmaker. SHE earned a masters and P, H, D.

From M, I, T. Her book to mal illusions, the false promise of optimization, was published last year. Her consulting business helps research teams with computational science and strategy in agricultural climate science, logistics and biosciences, SHE writes under linking profile.

I have some fancy pants degrees from fancy schools, and I was a professor for a while, but I don't consider myself a fancy person these days, he writes. I mostly live in a small role place where I get to pretend to be a farmer. In our conversation, coco and I discuss what is optimization, what are the benefits and costs.

We explore A I, and we are overlaps with optimization. We consider the present counts of optimizing our lives and finally discuss the extreme chAllenge of opting out. I hope you enjoy this year and episode will be back on january eighth, twenty twenty five with five or six.

Coco, welcome to money for the rest of us. I read your book this year. We touch on its some in an earlier episode. This book in the concept of optimization so intrigued me that I wanted to bring on the show and kind of have a conversation of the positives and negatives of optimization. But I look to start off, how do you describe what optimization is in some the benefits and ultimately some of the trade offs, which no clearly is which your books about. But just if somebody doesn't know what optimization is, what is IT?

okay. Well, we could start. There is great to be here, David. Thank you. So optimization, you know, i'd like to think about IT as a there's a technical definition and then there's sort of a cultural a cloquet definition.

And the technical definition is it's a set of technologies to, within a bounded space, set an objective function and and find the optimize, minimize or maximize some value within that space. And you know, I am former interested in the sort of cultural implications of the set of technologies that been evolving for a few centuries. But really, evolution as a technology in a way of thinking has accelerated in the last few decades.

And that cultural definition is, you know, IT takes its cue e from from the technical one. So it's minimizing or maximizing. And IT involves, you know, a way of thinking that constrains things in order to minimize or maximize some value. So we might be talking about maximizing or minimizing downtime in transportation schedules in order to do that.

To make that optimization involves a huge set of philosophical assumptions, right? That in this case were or minimizing time, right? Or a minimizing wasted time and were arranging things um and looking at the world, looking at schedules and looking at airplanes or trains or whatever whatever schedule looking at in in a particular way that allows for that optimization to happen and and that sort of where the where the book know kicks off from is is trying to understand how that point of you came about.

What are some things like clear benefits?

Where is to let us there are so many benefits all all round, right? And I often get push back or or accused accusations on this book of of being a lot of being against top timide ation. And i'm actually totally blown away by many of optimization fruits.

The book is is simply uh, an inquiry into what are the the tradeoff associated with with those fruits in. And so what we've lost, but as some examples, right, I go into modern agriculture in the book, we have wider ray of food stuff s at and I wrote this or did the research and before this latest wave of inflationary pressures. But i'd still argue it's it's largely the case that, that we have Green era of some of the cheapest food in human history and IT just the abundance in in terms of quantity and availability, variety of of food stuff s around the world is mind boggling.

And that is a result in large part of optimization of all sorts, right, treating farming as something to make more efficient has has let us hear certain monetary inventions. Commodity markets, for example, has let us here certain optimization, as I is looking into earlier in in transportation or or supply chains in getting patrol and to different places on the globe has allowed that. So so that just one example is the cheapest and availability of food. But there are numerous others, you know, most of them pertaining to kind of the the quality and ease of our western lifestyles.

Before we continue, let me post and share some words for one of this week sponsors net sweet. What does the future hold for your business? Ask nine experts and you'll get ten answer. Ball market, bear market rates rise, fall, inflation, upper, down. Be great if we had a Crystal ball, but we don't.

Until then, over thirty eight thousand businesses have future proved their business with net sweet by oracle, the number one cloud E R P, bringing accounting, financial management, inventory, hr into one fluid platform with one unified business management sweet. There's one source of truth giving you the visibility and control you need to make quick decisions with real time insights and forecasting. You're peering into the future with actionable data.

When you're closing the books and days, not weeks, you're spending less time looking backwards and more time on what's next. I know as our business growth, we'll certain tainted consider using net weet by oracle. Now whether your companies aren't millions or even one hundred of millions, next day helps you respond to immediate chAllenges and seize your biggest opportunities.

Speaking of opportunity, download the CF hos guide to A I A machine learning at net sweet. That comes last, David. The guide is free to you at net sweet dot.

Come flash David. Net sweet dot. Come slash David.

So if that's the case, then we should just keep optimizing and optimizing what what's bad about IT.

right? Well, yeah, I obviously if if I believe that, then then I wouldn't have written the book if you know I think just to keep building on the agricultural example, right, we've and depending on on what perspective you you're coming from, you know there planning of people who would argue that the quality of our food in terms of its you know nutritional value per calorie findings and minerals, both in the the end product and and in the soil has declined.

We are by many measures, less healthy in the years since commodity agricultural has accelerated. Then then we were before we another perspective would would have that the abundance or over abundance of food has has made us soft in in various ways, both literal and psychological. So so there's that perspective.

There's also the perspective that we've lost practices and and ways of being, ways of ways of farming that have all bit evaporated. I spend a lot of time in the lead up to to writing the book um with commodity farmers in in the midwest. I profile some sugar beat farmers in the decoder but i've also spent time with with corn and soybean farmers and you know all of them speak to just the over the the last few generations a distinction of, you know, their their profession from the people in their community, right?

IT IT used to be not so long ago that farmers would feed people nearby, or or at least people in the same nation. And that's no longer the case, especially for certain crops, and specially for crops that are not necessary, you know can be food stuff, but also can can go to feed animals or produce fuels. For example.

is there a way to have the best of both worlds? Or does this optimization ultimately, I guess, you're optimizing for something. So I me is is just one objective function.

So the cost is IT that just the mathematics of optimization can we optimize for more than one thing or is IT? It's several things at once. When when you talk about the supply chain or or something like that.

yeah and you know that you can optimize for whatever you want. And I think that's that's how how people use IT.

You know I smiled a little bit that you're phrasing because I I think we use that idea increasingly colloquially, especially out of the mouse of business school students, right? What are you optimising for? Are you optimising for happiness or the incommode, whatever IT may be? This is one of the questions I I get a lot, especially from business school types or or economist types. Rate is well, then if if we're not optimizing for the right things, why don't we just switch what were optimizing for? And I think that's that's a very valid response.

But to me, IT IT shows how embedded and entrenched this idea of optimization is, right? If the response to will our models, our modeling is falling short, is what we need to improve the models, right? IT IT belies a kind of a foreignness with thinking about what, what if we saw the world through different lens, then shoving IT into to a model and and trying to minimize or maximize one thing or or multiple things, whatever.

in the case, maybe, why is this so difficult to undo an optimization? You get example, agriculture. Or if the harm turns out to be greater than the benefit, why is an optimization and done more? Or in other examples of that?

Yeah I mean, I think yeah there there are examples of an undoing optimization, right? I I think there are maybe a partner when I I don't know this is a perfect example because I I haven't thought about IT, but I just off the cuff, right? It's sort of these large language bottle are are disuse and with a lot of these kinds of models rights, they don't perform as well until you program in A A certain amount of noise or randomness.

And I think again, it's not perfectly analogous. But when you are programing in right twain, an optimization to the optimize, an example with scheduling, for example, would be programing in scheduled downtime. So and with supply chains to or we're doing simulations to sort of account for, for those scenario, those are examples of programmatically d optimizing that, that have been successful. So again, I think there's one line of inquiry is how do we make optimizations Better within the framework of optimization? And that's that's a super interesting line of inquiry.

I guess my kind of my line of inquiries is really know what are the implications of our reliance and optimization in in so many domains in our in our lives and and kind of tracing that that history are, trying to understand how we got here, and how does this way of thinking will they continue to to have legs, right? We see seeing a lot of breakdowns in over the optimize systems. And to me, that makes me think, well OK this this idea of optimization is a it's a coherent metaphor right now, and we cycle through metaphor is right, like know the brain is a computer or or the heart is a hydrology pump, depending on what technologies where we're currently engaged with. And just makes me wonder what is know how much longer does this particular metaphor have have legs for and and what comes next?

So is what an example of, and something that over optimization and in its broken down.

I use IT a few examples in the book. We saw, you know, a lot of these. I think the early days of the code pandemic bright some of these to the four. So trying to your having supply chains that are very intricate and tightly wound and such that one you know, the breakdown of with one ship, one cargo ship getting stuck can have implications for weeks and weeks on end for for millions of of people. So if we weren't so reliant on those particularly optimized supply chains, right, we would have found substitutes and doing things and products that were more local, perhaps, and wouldn't have had that kind of fragile ity that that one cargo ship could throw things to thunder for so long.

So with an optimization like with with the cargo ship, like an L A member, I think an Operation to managment clash and graduate school and you know talk about bottle next. And is there a way to to optimize that? It's more. Distributed, right? So in the supply chain that mean that china minimize costs.

So there was just one way to do, but IT seems like there would be in a more complex optimization to the example gave optimized or putting some downtime in or IT is make IT more, the whole process is making more diverse, less chance to break down that in of itself. That is an optimization. correct?

No, correct. And people, all these technologies are, are super sophisticated. People are doing that, right? IT is, I guess my point is, is simply that if we hadn't made certain trades right over time, which which is generally convenience and Price or or lower costs um a desire for for more more variety. If we hadn't made various of those trades over time and we'd made the choice, say continued to school eze lower Prices out of by by you, creating things in factories abroad and shipping them over here, we would have maybe different kinds of fragilities. But but I argue we'd have not as fewer fragilities in, in a less optimized system even if, to your point, we're doing those optimization in a in a very smart and sophisticated way.

Before we continue, let me pause and share some words from this week. Sponsors, we've talked previously about how much of your personal data is out there on the internet for anyone to see your name contacting, vote, social security number and home address.

Even information about your family members are being compiled by data brokers and sold online as a person who has a public presents on incredible aware about safety and security and how privacy matters more than never. Now, while I use delete me and recently gave a gift of delete me to my songs, and they've been excited to get their their first privacy report showing what data had been removed from the inter net. With delete me, you can sign up and provide them exactly what information you want to lead IT and their experts will take IT.

From there, you'll get a personalized private report like I just received, showing what they found, where they found IT and what they removed. That's because delete me isn't just a one time service. There are always working for you, constantly monitoring and removing the personal information you don't want on the internet could simply delete me, because all the hard work of wiping your and your family's personal information from data brokers, websites. So take control of your data and keep your private life private by signing up for delete me now with a special discount for our listeners, ers and their loved ones today. Get twenty percent off your delete me plan when you go to join delete me that comes less David, and use promotion de David twenty and check out the only way to get twenty percent off is to go to join delete me god comes less David twenty and enter code David twenty at checkout that join delete me don't come slash David twenty cold David twenty .

so an article today. I think when dees was used, I guess, to have a independent corp device, all the supplies and they they're gonna use A I now supposedly in their supply chain optimization for the picture, don't run out of frosty if they're doing in a dollar special. We've an absolute N A.

I and I have A A decent understanding of least large language models. But in an article like that makes a sound like this could be magic, we're gonna turn IT over the a it's going to figure out whether that is enough frost in the morning, you know, at some specific restaurant. And we talk about algorithms like how is A I working with optimization? Because, Franklin, my experience in using large language models is there they're great tools. But like I wouldn't trusted to actually make a real, real decision that was important yeah.

So I don't know I haven't read the the article about wind's, so I don't know how there.

but just in general like that and because that was always maybe the same to refer the question how how does optimization plan to A I as A I is developing?

Yeah that's I don't know how to answer that question because people mean so many things when they say A I. And you know most of the time it's a term that's that's spandy about um IT IT can mean anything from kind of smart assistance to actual physical automation robots to sort of simply machine learning or statistical models that that have been around for for a while already. So I guess your question is how do how do A I an optimization in our sect? I think I would be helpful .

to to kind of you how are they relate? I mean, we are overlap. I mean, if you have an optimization function in this case, I know what the details either because they never shared them, but they made IT sound like, okay, we have a supply chain which is optimized and his human elements.

And then they say, well, were bringing in this A I to make a lot of these decisions. They almost make IT seem like IT is panache, where IT seems like it's not that simple, right? In that I guess overlap in terms of A I and optimization.

Yeah so one point I can make us suppose that you know a lot of these AI models or you know different class of A I models have minimization or maxims ation functions built into how they are working. No, that's not exactly what you are asking. But um you asking more, how does A I get sort of layer on top of optimization? You know i'm kind of i'll show my two colors here.

I'm neither A A I boomer nor nor do mor. I'm kind of just a met like I think the I think we're in a very large bubble. We have a lot of very intelligent people working on A I and a lot of money being thrown at IT.

And it's unclear to me. I mean, obviously, there are some important use cases and some important efficiencies that, let's say, large language models in particular, going to a help with and solve or or begin to solve. I I don't know that those efficiencies and then the use cases add up to all the guggle ans of dollars that have been poured into .

this this industry question.

So yeah, I mean, I I think there's there's some overlap on how some of these models are are working in, in terms of optimizing certain things along the way. And I think there's also over in, in terms of A I you know some of these newer developments can be applied in in different ways to to supply chains. For example, like I don't know that there's like A A huge use case there for large language models in particular, just because what what those technologies sell ad is going through large corporations of of taxi doing various tricks they are in.

So so IT is optimization. I think what we're getting out there is a lot of terms Brown around optimize A I in at the narrows optimization. Is this mathematical function where you're trying to make in the case of finance, on a portfolio theory, we have to expect to returns.

We have volatility assumptions with correlations you were trying to minimize for given level return. We're try to minimize the volatility for a level volatility or trying to maximize the return. Is that what we mean when we say optimization at a simple is just in a program is just that mathematical function? Yeah okay.

So it's it's very new. yes. And then so like a supply chain or agriculture, in this case, the mono portfolio theory since principle in its garbage in garbage out itself is pretty straight forward. Because, you know, when I ban as money or and is to survivor, than you would put all these constraint so you would get the answer you wanted. Is that pretty common? When IT comes to optimizations, they put a bunch of constraint sons, so they get the answer they want or they can supply chain or did he is a tweet a lot typically in practice and other other domains.

So if we're talking about linear programming in particular, I mean, I think supply chains and transportation energy, those are actually great use cases for for lennar programing, right? Because you know where we're dealing with concrete, very measurable parameters, right? Often take my time or money distance volumes, right? Those are very memorable things.

And we have discrete outcomes. We know what we want to optimize. We are and our we have a lot of data, right? The evidence for we could get run experiments and see the outcomes and and know know how we did, right?

So the the results are right there, know an argument financed and things like, you know a lot of the the tech of search, for example, you know those those actually aren't aren't bad applications either, right? Like you say, there's the garbage, garbage out there may be more garbage when you're when you're talking about information. And there are all kinds of reasons why in the the imperfect world in which we live, modern porfolio theory fails.

But in a, in artificial simulated world, or even in a world like online advertising, where you have lots and lots of data, some of these technologies actually work. They work pretty well, right? I don't know. I mean europe, maybe you disagree, like I don't know about finance, but oh.

no, no, I mean, I I guess you know maybe i'm overreaching thinking and I think he comes down to what optimization means in the sense that you this narrow mathematical formula we're trying to optimize, then it's supplied everything else. So like Marvin king was a central banker for the bank of england. Once humans we don't optimize, we cope.

And I think there's IT seems like there's this tendency to apply optimization to everything, including our our day to day life. And IT seems like your book touched on that a little bit in kind of paradox like you like sometimes the world seems so optimize. We wanted escape and and just not do IT at all. And IT is IT more a case of were miscible ying optimization or IT isn't really optimized in a mathematical way. Do you want to get that IT IT just IT seems that we have taken these terms like optimization A I and we missed IT because it's more of a narrow thing and then we try to optimize our lives, which in fact is very difficult because this is not one objective function.

Yeah I I think i'd um agree with you on the diagnosis, maybe how nip c the the reasoning right with IT, which is I don't think so.

I think the problem with optimizing our daily lives is not that there's not just one objective function is that this way of thinking is is sort of creates the the spiral, right? And I see a lot of people who, you know, I think I use the word canvas zing in in the book, right? Like like once you start to optimize, you want to optimize more.

And we forget that maybe our daily lives don't need to be super efficient, right? It's not just that we might want to optimize for for many things in our lives, right? Money, free time, joy, right? But but also that. You know, some of those things that we care about, our our values don't even lend themselves optimization. They are not money or or free time or things that are that .

are that measure that s so in your book you mention what are the attention between seeking control through optimization and and just wanting to to escape? Take a break from IT at center. Where has you're thinking evolved at this? Because at least at this, as I was reading a book that sound like you are, sort of you do have a conclusion of answer like where where's the right baLance between optimization and and stepping back and either d optimizing or escaping?

Yeah and I still I I don't have a good answer right now. A lot of readers who no, I mean, I am not saying IT frustrate to you, but I think IT does frustrate.

I've gotten that feedback that what why is why? Why aren't there more conclusions, right? Or or take away like a checklist of ten things you can do and and I think that's a fair criticism, right? I think if I had to sort of states simply IT take away general, very general take away right IT would be sort to consider process over outcome, to consider specifics over abstractions, to consider going going more slowly or slowing down.

And that said, I think it's I think this tension between wanting to optimize and wanting to opt out is is a productive one, at least for me, right? I I worked for for a long time in know that sort of the peak of an era in in data science. And I am still fascinated by the power of these these technologies. I also feel overwhelmed sometimes I just tell how quickly the the modern world moves and um how many choices we have to make on a on a daily basis and and how ill equipped we are to or I am, I should say, to to make those choices given the the intellectual framework and and moral moral and intellectual and cultural frameworks that we have today. So yeah I mean and I think you know when you're writing a book, you're sort of playing a character to write and and it's the the character that that I am in the book and represents me and represented me when I was writing IT, right is is a character that evolving from this place of of being an optimization to trying to figure out why i'm that way and and maybe what what are some alternatives maybe that.

So i'd like the fact that there weren't bull points at the end. I mean, I prefer books that kind of that wonder in weave. So the book came out here.

So goes, have you made additional stride? Degas, in terms approached the light, in in terms of kind, resolving or managing the attention? I mean.

you well, clearly not because because we're we're here with the the got my head set on in the modern internet, so haven't become a long gun a mountain yet. But you know, is that where where I think I was headed in in the book, right? I I I hope I make this point clear, right? I don't think it's possible or or to opt out today is an an incredibly extreme act, right? And and very, very difficult right.

Even you know living in a remote real place like i'm connected and tidy and in so many ways, rates, not least supply chains, the internet um ways of thinking, communications, all these things so I wouldn't say i've opted out in in any way. I don't think it's possible for me at least I think it's if extreme stoic and in the old old school sense, right maybe so. Has my thinking evolved since then? I not not in a major way. I think i'm continued to be amused by all the attention that's that's being paid to A I or are these these kinds of um new technologies um and I think in in many ways that there there will sort of a self filling propac's to IT that that there are so much money in IT IT IT has to yield yield so much money in so many smarts and IT I should add that that IT has to yell something so yeah I am curious if how you're thinking somebody in finance. And now maybe I don't know how would you describe your relationship ended to finance, but how you're thinking about optimization is evolved.

I try to avoid that basically. So I an in in finance, specifically mono portfolio theory, investors don't think in terms of the risk as volatility. They they look at them as losing money.

And and and so you know our models is on how much could you lose and not be ruined and in complete running out of money. And so i've tended to back away from optimization and build in a lot of space in in, in slack, in redundancy. That just appeals more to me.

No, not an up to you know, I stop counting steps, for example, or or trying not to look at the APP and wrapping up then just back on A I, I, I agree IT seems like there is a lot of money in like we use IT IT helps us. But on the other hand, and IT IT just like is not taking over the world by any means to right to think of anything. Hopefully IT will increase productivity.

Is that what you're seeing? Or is that just like IT sounds like you're skeptical how much money has been put into IT and he has to yield something? Do you there's a lot of fear monger when IT comes to AI. Are you afraid of where he is going? Is some breakthrough as you know, some of the people in the space or are fearful you you take IT sort of overhit ed, it's really good at predicting words and perhaps some other things, but it's not taking over the human race.

Yeah I mean, like I said, I am maybe on a in a camp of one here, but I am neither a boomer nor nor a duomo. So I don't think no I think IT will cause some disruptions. I I don't think is is um a new life form that's gonna take over in the world out out.

These models are used in such a limited space, right? We we think it's unlimited because we're so enraptured by information right now and then the internet and and it's important riders. It's how we know.

It's how we understand the world. But I I don't think, number one, the technologies are good enough and more importantly, and they're quite good, but they're not good enough and and more importantly, they exist in such a limited domain. They don't there's no physicality to them.

They don't they're not living things um with know with cells that replicate and you know means there's all this attention on me, means aren't aren't genes right there. There are different things there there there's um they exist in a space of abstraction. So I I think they will cause disruption, but I don't think they're in the world and I also don't think these models are going to save the world, right?

I think and this is a little bit condescent or demeaning to to call IT sort of a glorified clippy. But I think and I but I think those are the domains, the domains of information where these models, which are far more sophisticated than than clippy, will have an impact and theyll be moderately disruptive there. There will be jobs lost, right? There will be industries there turned on their heads. But I I yeah repeating myself, but I don't I don't think they're are .

gonna be in the world. No, no, you're not. I I agree with you. I mean, from what i've seen that I think the overpromising often times is because they have to put such high valuation on the companies and how to raise money.

So of course, yeah, but when you see I can practice is use IT try to use IT. Some days it's like this is great and some days this thing is just stupid. I think because it's it's all statistics. S well, okay, just a rap of what type of projects are you're working on now? Are you fighting interesting how you're working in another book or what are you spending your time these days?

Oh gosh, a lot of things, some things, family stuff. I do have a couple writing projects, not their very licence stages or not. Not ready to talk about specifics there and doing a little bit of of teaching and consulting still. So a whole, a whole bundle of things, but most the mostly just living.

they go wonderful that is there. If readers want to go to hold to you a website you wanted share, buy your book, as I have done, or my .

publisher would love IT and as what I, they would buy my book. And I do have a website, coco folio dot com. awesome.

great. Thank you, coco.

Thank you so much.

I hope you enjoy this conversation with coco chrome again. Her book is optimal illusions to false promise of optimization. Her website is coco folio dot com that C O C O F O L I O dot com. Thanks for listening. You may be missing some of the best money for the rest of our content.

Our weekly insider guide aim on this letter goes beyond what we cover in our podcast episodes and helps elevate your investment journey with information that works best in written in visual formats. With the insider sy IDE, you can discover investing in economic insight provided only to our news letter subscribers, unlike greater investing confidence with high value snipers from our premium products plus membership and asset cap, further connect with the money for the restless team and community. And when you sign up, will also send you our exclusive investing checklist to help you invest with more confidence right away.

You'll also get our introductory email series on eight essential investment principles that IT followed, can make you a Better investor, will also share our recommendations for podcast episodes. Ticks in books, the insiders guide is the best next step to get the most out of your investment journey. If you're not on the list, go to money for the rest of us that come and subscribe right there on the home page.

Everything i've shared with you in this epsom been for general education. You've not considered your personal risk profile, if not provided investment advice. This is simply general education on money investing in the economy. Have a great holiday season.