In the ever-evolving narrative of technological advancement, artificial intelligence (AI) stands out as a particularly profound development, reshaping industries and investment paradigms alike. My own journey with AI, beginning with a simple artificial hand crafted in my college days, mirrors the broader trajectory of innovation and its implications for the financial sector.
This exploration into AI’s role within the investment world is not merely an academic exercise but a critical inquiry into how we must adapt our strategies and mindsets in the face of unprecedented change. The nuances of AI's impact on finance, from automating routine tasks to redefining the essence of strategic insight, should be viewed in the broader context of understanding cycles, recognizing patterns, and anticipating shifts within this dynamic landscape.
This narrative seeks to offer observations derived from the intersection of technology and investment philosophy, guided by the belief that the most significant advancements—and the most remarkable opportunities—often lie at the confluence of change and adaptation.
The Artificial Hand
My foray into AI began with an undergraduate project almost three decades ago: a rudimentary artificial hand crafted from conductive foam used to ship integrated circuits. Inspired by an article about the foam’s conductive properties being proportional to pressure applied, I programmed the hand (in assembly language) to determine how much pressure to apply depending on the object it was holding (e.g. apply less pressure if holding an egg). It was artificial, and the intelligence was limited.
I learned about algorithms such as simulated annealing and genetic algorithms, and how they could be used in optimization problems. Growth at that time was linear and isolated- dot com was yet to see its bust, and a ‘modem’ was still a thing.
My Transition to Finance
After college I joined the derivatives desk at JPMorgan and my career turned to finance.
While I continued my journey into the world of quantitative finance and investing, managing portfolios for pension plans, and helping build and scale investment business, the seeds for the AI revolution were being sown elsewhere. Like the beanstalk from Jack’s magic beans, from these seeds AI would grow at a rapid and astonishing pace. This AI journey is well known - the internet, superior computing, open source compounded growth and provided motivated individuals and organizations the raw tools needed to scale the workhorses of AI, namely machine learning (ML) and large language models (LLM).
Early cousins of what is today considered AI continued to play a role as I made my way through the investment world. First as a tool - “natural language processing” enabled non-technical team members query databases without knowing SQL.
Then as an investment tool - expert systems helped systematize and reduce the impact of human emotions (and by extension reduce mistakes) in investment processes. I was fortunate to be asked to become a partner at a startup investment manager that was using AI and machine learning techniques to democratize hedge fund strategies. At Voya, I became the first portfolio manager to include AI strategies developed by an investment team that the firm acquired.
The Evolution of AI: A Quantum Leap
All this pales in comparison to the progress AI has made in less than 400 days since ChatGPT came to market. While AI adoption was high even before ChatGPT came on scene, its frictionless user experience brought AIs capabilities out of the shadows and into everyone’s consciousness.
Let me share a few of my recent experiments with Generative AI and then zoom out and discuss some of the broader implications.
AI based assistants, from ChatGPT, the swiss-army knife of GenAI, to hundreds of ‘purpose built’ applications available, are transforming the way work gets done. I have been using ChatGPT to help me speed up writing code and analyze data, and its remarkably good.
The screenshots below are from an interaction I had, where I wanted to see if chatGPT could create a mathematical model for me based on input data. Even when it failed it was useful. It provided suggestions on how I could run the code in my own environment, and even provided me with processed data to feed my code.
For those interested, here is the output of the code when I ran it on my side:
GenerativeAI can also transform marketing. Market outlooks, commentaries, and even intelligent ‘strategists’ can be created by having AI write the commentary and then apps like HeyGen have an AI avatar share that strategy. Here is a fascinating (and creepy) video of my HeyGen avatar talking about the markets. It took me a total of less than 10 minutes from start to finish to create this. I am both fascinated by its potential and scared by its potential misuse.
The Investment World: Redefined by AI
The implications for the investment industry are profound. A few thoughts on that below:
Reimagining Value: AI will make you question the areas where an investment manager truly adds value: If a ‘market update’ or investment commentary can be written by AI, and at a level where its hard to make out the difference between human commentary and AI commentary, then what is the value of that commentary? Do we end up in a world of “passive” AI commentary and then investment managers having to truly come up with something unique and different that is valuable to their investors.
Taking that further, it also makes the investor or allocator ask the question - how much of what I do and what I get from my investment managers is really adding value? What do I need and what can I sub-contract to AI?
Creating a Better Investor: Can AI make you become a better investor by 1) using large language models to make sense of unstructured information, and then combine with structured data and apply tried and tested investment principles 2) reducing the barriers to analysis 3) Checking your own behavioral biases 4) Acting as your investment coach, helping you refine and improve your understanding
Commoditizing an undifferentiated Core: In an environment where the core work of investing becomes commoditized, what does it take to become a better investor? Can AI help investors better understand the difference between luck and skill? While there will always be luck involved in investing, perhaps AI may lead to identifying those managers with the right combination of luck and skill? This could help allocators identify those managers whose performance is driven by more ‘bankable’ factors, and less by idiosyncratic luck. This can lead to commoditizing an undifferentiated “core” of investing, while focusing on opportunistic alpha, portfolio construction, and risk management as main drivers of alpha.
A new breed of managers will thrive: Paradoxically, greater AI increases the importance of the human investor - except, this time it’s a different kind of individual or team than what allocators have traditionally gravitated toward. The value-add of a human analyst or portfolio manager will not come from the ability to process information, nor will it be from analyzing earnings reports and other kinds of activities investors presently ascribe value to. Instead, AI will unleash a new kind of human analyst, portfolio manager and CIO. One that is free from the narrow confines of their own technical capacity and instead can use their insights, frameworks, mental models, and mindsets to come up with a core thesis, which can then be implemented in multiple markets or areas that they were unable to access because of technical limitations.
One particular area with the greatest opportunity for such a revolution (that I am focused on) is bridging the gap between public and private markets. Humans with a keen understanding of markets and investment theory can use AI and related tools to transcend the barriers between public and private markets in order to significantly transform the investment experience for the investor (LP).
Beyond Finance: The Societal Implications of AI
The world is adopting AI at a rate faster than any other revolution that has come before. The Industrial Revolution took half a generation, the Internet took decades. GenAI seems to have entered mainstream use at what feels like the blink of an eye. The impact of AI will be enduring.
This is because not only does AI open up new avenues, it's also making us learn more about ourselves. For example, personalized AI coaches can help children study and understand course material better.
Jens Ludwig and Sendhil Mullainthan in their paper “Machine Learning as a Tool for Hypothesis Generation” provide a framework for how scientific inquiry can be advanced using Machine Learning to generate (not just test) hypothesis. They also share their troubling discovery that just pixels in a mugshot can explain more than half the variance in a judges decision on whether to incarcerate a defendant.
And in the search for improving artificial intelligence, we are learning more about our own intelligence, how it came to be, and more importantly why it came to be. Max Bennet’s book “A Brief History of Intelligence” is a fascinating journey into the evolution of intelligence. The more we progress in the world of AI, the closer the domains of biology, technology, philosophy and sociology are going to come together.
Ethical Considerations: Navigating the AI Landscape
I enjoy playing ‘Devil’s Advocate’, thinking of ways to challenge my views, and asking why the other side holds the views that they do. What are their set of facts, opinions and beliefs? Despite being a cheerleader for AI, and an early adopter, I’ve started to appreciate some of the issues that critics have brought to the forefront. My HeyGen avatar, while it can improve my marketing throughput, also shows the potential for misuse and disinformation. The more I rely on ChatGPT the more prone I am to replicating any error in it.
The standard of ‘trust’ for AI is higher than that of a human. A good friend of mine, who is an authority on data science, machine learning and AI, and a leading technologist, believes that the bar for AI to be ‘accurate’ is far greater than a human - while we may be more accepting of ‘human error’, for some reason, as humans we will always hold a machine to a higher standard. Which leads to the need to be able to “trust” AI.
The risks of AI go beyond just accuracy. Indeed, humans have lived with inaccuracies and imperfections in our own systems for generations. In a paper Managing AI Risks in an Era of Rapid Progress, authored by some of the leading thinkers in the industry such as Yuval Noval Harari, Dan Kahneman, and Ashwin Acharya have described some these risks and shared their opinion on what needs to be done (side note - Ashwin spent a few months as an intern for my fledgling startup in 2016 before moving on to bigger and better things - that’s my claim to fame!).
A Portfolio Optimization Approach to AI
While the risks of AI are real and need to be addressed, taking a portfolio view toward the risks and rewards of AI is needed.
In investing, we know that in order to maximize the odds of achieving our objective, we need to take risks. The key to successful portfolio management is not to avoid risk. Neither is it to take an unrewarded risk. Instead, successful investing is about exposing yourself to a broad set of risks, these risks are diversified and uncorrelated, and for which you get paid a risk premium, and the expected value on balance is positive.
Understanding the probability and severity of loss goes a long way in deciding what risks fit and maximize your utility function.
And for those risks that are unrewarded, or may lead to ruin, paying to hedge those risks, or having someone underwrite those risks are prudent.
Perhaps such a portfolio optimization approach is needed in our approach to AI. There may be some large systemic risks that need better governance and regulation, perhaps a set of standards are needed? (I am not an AI policy expert, nor do I want to be one).
At the micro level, as a user of AI, instead of using fear or insecurity as an excuse to shut down progress or innovation, perhaps the exercise is more about managing those risks in order to achieve something truly amazing.
Perhaps true creativity and innovation lies in understanding where AI is valuable, and where plain linear regression or human intuition is more suitable. And maybe that’s how an individual or an organization can make themselves valuable in the age of AI.
Conclusion: Charting a Course Through Uncharted Waters
The journey with AI, from its humble beginnings to its current prowess, mirrors the broader evolution of technology and society. As we stand on the brink of this new frontier, the lessons learned from the past—both in technology and investing—serve as valuable guides. Embracing uncertainty, fostering innovation, and maintaining a balanced perspective on risks and rewards will be key to navigating the future. AI, with all its potential and pitfalls, offers a unique opportunity to redefine our approach to investing, decision-making, and, ultimately, our understanding of the world.
(The introduction, conclusion and headings were written with the assistance of ChatGPT)
Not Investment Advice. Not an offer to buy or sell securities or investments. For Educational Purposes Only. Reflects my personal views. I attest am human, though my daughter says that’s sometimes questionable ;-)