Beat your Bot: Building your Moat against AI
The Bots are here...
It seems like a lifetime has passed since artificial intelligence (AI) became the market's biggest mover, but Open AI introduced the world to ChatGPT on November 30, 2022. While ChatGPT itself represented a low-tech variation of AI, it opened the door to AI not only as a business driver, but one that had the potential to change the way we work and live. In a post on June 30, 2023, I looked at the AI effect on businesses, arguing that it had the potential to ferment revolutionary change, but that it would also create a few big winners, a whole host of wannabes, and many losers, as its disruption worked its way through the economy. In this post, I would like to explore that disruption effect, but this time at a personal level, as we are warned that we risk being displaced by our AI counterparts. I want to focus on that question, trying to find the middle ground between irrational terror, where AI consigns us all to redundancy, and foolish denial, where we dismiss it as a fad.
The Damodaran Bot
I was in the eleventh week of teaching my 2024 spring semester classes at Stern, when Vasant Dhar, who teaches a range of classes from machine learning to data science at NYU's Stern School (where I teach as well), and has forgotten more about AI than I will ever know, called me. He mentioned that he had developed a Damodaran Bot, and explained that it was an AI creation, which had read every blog post that I had ever written, watched every webcast that I had ever posted and reviewed every valuation that I had made public. Since almost everything that I have ever written or done is in the public domain, in my blog, YouTube videos and webpage, that effectively meant that my bot was better informed than I was about my own work, since its memory is perfect and mine is definitely not. He also went on to tell me that the Bot was ready for a trial run, ready to to value companies, and see how those valuations measured up against valuations done by the best students in my class.
The results of the contest are still being tabulated, and I am not sure what results I would like to see, since either of the end outcomes would reflect poorly on me. If the Bot's valuations work really well, i.e., it values companies as well, or better, than the students in my class, that is about as strong a signal that I am facing obsolescence, that I can get. If the Bot's valuations work really badly, that would be a reflection that I have failed as a teacher, since the entire rationale for my postings and public valuations is to teach people how to do valuation.
Gauging the threat
In the months since I was made aware of the Damodaran Bot, I have thought in general terms about what AI will be able to do as well or better than we can, and the areas where it might have trouble. Ultimately, AI is the coming together of two forces that have become more powerful over the last few decades. The first is increasing (and cheaper) computing power, often coming into smaller and smaller packages; our phones are now computationally more powerful than the very first personal computers. The second is the cumulation of data, both quantitative and qualitative, especially with social media accelerating personal data sharing. As an AI novice, it is entirely possible that I am not gauging the threat correctly, but there are three dimensions on which I see the AI playing out (well or badly).
Mechanical/Formulaic vs Intuitive/Adaptable: Well before ChatGPT broke into the public consciousness, IBM's Deep Blue was making a splash playing chess, and beating some of the world's greatest chess players. Deep Blue's strength at chess came from the fact that it had access to every chess game ever played (data) and the computing power to evaluate 200 million chess positions per second, putting even the most brilliant human chess player at a disadvantage. In contrast, AI has struggled more with automated driving, not because driving is mechanically complicated, but because there are human drivers on the surface roads, behaving in unpredictable ways. While AI is making progress on making intuitive leaps, and being adaptable, it will always struggle more on those tasks than on the purely mechanical ones.
Rules-based vs Principle-based: Expanding the mechanical/intuitive divide, AI will be better positioned to work smoothly in rules-based disciplines, and will be at a disadvantage in principle-based disciplines. Using valuation to illustrate my point, accounting and legal valuations are mostly rule-based, with the rules sometimes coming from theory and practice, and sometimes from rule writers drawing arbitrary lines in the sand. AI can not only replicate those valuations, but can do so at no cost and with a much closer adherence to the rules. In contrast, financial valuations done right, are built around principles, requiring judgment calls and analytical choices on the part of appraisers, on how these principles get applied, and should be more difficult to replace with AI.
Biased vs Open minded: There is a third dimension on which we can look at how easy or difficult it will be for AI to replace humans and that is in the human capacity to bring bias into decisions and analyses, while claiming to be objective and unbiased. Using appraisal valuation to illustrate, it is worth remembering that clients often come to appraisers, especially in legal or accounting settings, with specific views about what they would like to see in their valuations, and want affirmation of those views from their appraisers, rather than the objective truth. A business person valuing his or her business, ahead of a divorce, where half the estimated value of that business has to be paid out to a soon-to-be ex-spouse, wants a low value estimate, not a high one, and much as the appraiser of the business will claim objectivity, that bias will find its way into the numbers and value. It is true that you can build AI systems to replicate this bias, but it will be much more difficult to convince those systems that the appraisals that emerge are unbiased.
Bringing this down to the personal, the threat to your job or profession, from AI, will be greater if your job is mostly mechanical, rule-based and objective, and less if it is intuitive, principle-based and open to biases.
Responding to AI
While AI, at least in its current form, may be unable to replace you at your job, the truth is that AI will get better and more powerful over time, and it will learn more from watching what you do. So, what can we do to make it more difficult to be outsourced by machines or replaced by AI? It is a question that I have thought about for three decades, as machines have become more powerful, and data more ubiquitous, and while I don't have all of the answers, I have four thoughts.
Generalist vs Specialist: In the last century, we have seen a push towards specialization in almost every discipline. In medicine, the general practitioner has become the oddity, as specialists abound to treat individual organs and diseases, and in finance, there are specialists in sub-areas that are so esoteric that no one outside those areas can even comprehend the intricacies of what they do. In the process, there are fewer and fewer people who are comfortable operating outside their domains, and humanity has lost something of value. It is the point I made in 2016, after a visit to Florence, where like hundreds of thousands of tourists before me, I marveled at the beauty of the Duomo, one of the largest free-standing domes in the world, at the time of its construction.
The Duomo built by Filippo Brunelleschi, an artist who taught himself enough engineering and construction to be able to build the dome, and he was carrying on a tradition of others during that period whose interests and knowledge spanned multiple disciplines. In a postright after the visit, I argued that the world needed more Renaissance men (and women), individuals who can operate across multiple disciplines, and with AI looming as a threat, I feel even more strongly about this need. A Leonardo Da Vinci Bot may be able to match the master in one of his many dimensions (painter, sculptor, scientist), but can it span all of them? I don't think so!
Practice bounded story telling: Starting about a decade ago, I drew attention to a contradiction at the heart of valuation practice, where as access to data and more powerful models has increased, in the last few decades, the quality of valuations has actually become worse. I argued that one reason for that depletion in quality is that valuations have become much too mechanical, exercises in financial modeling, rather than assessments of business quality and value. I went on to make the case that good valuations are bridges between stories and numbers, and wrote a book on the topic:
At the time of the book's publication, I wrote a post on why I think stories make valuations richer and better, and with the AI threat looming, connecting stories to numbers comes with a bonus. If your valuation is all about extrapolating historical data on a spreadsheet, AI can do it quicker, and with far fewer errors than you can. If, however, your valuation is built around a business story, where you have considered the soft data (management quality, the barriers to entry), AI will have a tougher time replicating what you do.
Exercise your reasoning muscle: I have never been good at reading physical maps, and I must confess that I have completely lost even my rudimentary map reading skills, having become dependent on GPS to get to where I need to go. While this inability to read maps may not make or break me, there are other skills that we have has human beings, where letting machines step in and help us, because of convenience and speed, will have much worse long term consequences. In an interview I did on teaching a few years, I called attention to the "Google Search" curse, where when faced with a question, we often are quick to look up the answer online, rather than try to work out the answer. While that is benign, if you are looking up answers to trivia, it can be malignant, when used to answer questions that we should be reasoning out answers to, on our own. That reasoning may take longer, and sometimes even lead you to the wrong answers, but it is a learned skill, and one that I am afraid that we risk losing, if we let it languish. You may think that I am overreacting, but evolution has removed skill sets that we used to use as human beings, when we stopped using or needing them, and reasoning may be next on the list.
Let your mind wander: An empty mind may the devil's workshop, at least according to puritans, but it is also the birthplace for creativity. I have always marveled at the capacity that we have as human beings to connect unrelated thoughts and occurrences, to come up with marvelous insights. Like Archimedes in his bath and Galileo under the apple tree, we too can make discoveries, albeit much weighty ones, from our own ruminations. Again, making this personal, two of my favorite posts had their roots in unrelated activities. The first one, Snowmen and Shovels, emerged while I was shoveling snow after a blizzard about a decade ago, and as I and my adult neighbors struggled dourly with the heavy snow, our kids were out building snowmen, and laughing. I thought of a market analogy, where the same shock (snowstorm) evokes both misery (from some investors) and joy (on the part of others), and used it to contest value with growth investing. The second post, written more recently, came together while I walked my dog, and pondered how earthquakes in Iceland, a data leak at a genetics company and climate change affected value, and that became a more general discourse on how human beings respond (not well) to the possibility of catastrophes.
It is disconcerting that on every one of these four fronts, progress has made it more difficult rather than less so, to practice. In fact, if you were a conspiracy theorist, you could spin a story of technology companies conspiring to deliver us products, often free and convenient to use, that make us more specialized, more one dimensional and less reason-based, that consume our free time. This may be delusional on my part, but if want to keep the Damodaran Bot at bay, and I take these lessons to heart, I should continue to be a dabbler in all that interests me, work on my weak side (which is story telling), try reasoning my way to answers before looking them up online and take my dog for more walks (without my phone accompanying me).
Beat your bot!
I am in an unusual position, insofar as my life’s work is in the public domain, and I have a bot with my name on it not only tracking all of that work, but also shadowing me on any new work that I do. In short, my AI threat is here, and I don’t have the choice of denying its existence or downplaying what it can do. Your work may not be public, and you may not have a bot with your name on it, but it behooves you to act like there is one that tracks you at your job. As you consider how best to respond, there are three strategies you can try:
Be secretive about what you do: My bot has learned how I think and what I do because everything I do is public - on my blog, on YouTube and in my recorded classes. I know that some of you may argue that I have facilitated my own disruption, and that being more secretive with my work would have kept my bot at bay. As a teacher, I neither want that secrecy, nor do I think it is feasible, but your work may lend itself better to this strategy. There are two reasons to be wary, though. The first is that if others do what you do, an AI entity can still imitate you, making it unlikely that you will escape unscathed. The second is that your actions may give away your methods and work process, and AI can thus reverse engineer what you do, and replicate it. Active investing, where portfolio managers claim to use secret sauces to find good investments, can be replicated at relatively low cost, if we can observe what these managers buy and sell. There is a good reason why ETFs have taken away market share from fund managers.
Get system protection: I have bought and sold houses multiple times in my lifetime, and it is not only a process that is filled with intermediaries (lawyers, realtors, title deed checkers), all of whom get a slice from the deal, but one where you wonder what they all do in return for their fees. The answer often is not rooted in logic, but in the process, where the system (legal, real estate) requires these intermediaries to be there for the house ownership to transfer. This system protection for incumbents is not just restricted to real estate, and cuts across almost every aspect of our lives, and it creates barriers to disruption. Thus, even if AI can replicate what appraisers do, at close to no cost, I will wager that courts and accounting rule writers will be persuaded by the appraisal ecosystem that the only acceptable appraisals can come from human appraisers.
Build your moat: In business, companies with large, sustainable competitive advantages are viewed as having moats that are difficult to competitors to breach, and are thus more valuable. That same idea applies at the personal level, especially as you look at the possibility of AI replacing you. It is your job, and mine, to think of the moats that we can erect (or already have) that will make it more difficult for our bots to replace us. As to what those moats might be, I cannot answer for you, but the last section lays out my thinking on what I need to do to stay a step ahead.
Needless to say, I am a work in progress, even at this stage of my life, and rather than complain or worry about my bot replacing me, I will work on staying ahead. It is entirely possible that I am embarking on an impossible mission, but I will keep you posted on my progress (or absence of it). Of course, my bot can get so much better at what I do than I am, in which case, this blog may very well be written and maintained by it, and you will never know!
YouTube Video
Blog Posts (referenced)
Brilliant article ! Reminds me of a fast paced thriller where the hero managed to stay ahead of « the killer on the move » at times with only a few minutes of advantage …. .. hope the bot looses
I'm pretty sure this blog was written by the Damodaran bot