Ben Graham Value Conference IV coverage continues – stay tuned for more – Devin Haran compiled these notes on the talk by Protege Partners at the conference.
Date: June 20, 2017
Organizer: CFA Society of New York
Location: Fordham University Law School
Topic: “Can Man & Machine Beat Mr. Market?”
Also see
Kolanovic’s Deep Dive Into Artificial Intelligence Lacks Human Understanding
Speaker: Michael Oliver Weinberg, CFA (Protege Partners – Chief Investment Strategist)
Protege Partners Makes Another Wager
Last month, Protege Partners conceded that their renowned million-dollar bet with Warren Buffett was unlikely to pay out. Now, the $2 billion, New York-based fund of funds is ponying up again, this time backing computers in the rapidly intensifying contest between man and machine.
Nearly a decade ago, in the midst of the global financial crisis, Ted Seides, the former CIO of Protege Partners, placed a very public, million-dollar bet on behalf of his firm. His counterparty in that bet was none other than Warren Buffett. Seides wagered that over the course of the subsequent decade, a bespoke portfolio of 5 funds of funds, would outperform a low-cost Vanguard index fund selected by Mr. Buffett. With just months remaining until the wager’s deadline, Mr. Buffett finds himself in the lead, by what is effectively an insurmountable margin. Since the two parties shook hands some 10 years ago, the index fund selected by Mr. Buffett has delivered a 7.1% compound annual return compared to a 2.2% rate of return for the portfolio of funds of funds.[1] In light of this fact, Mr. Seides took the honorable approach and admitted defeat, conceding on behalf of Protege Partners in May.
Though Mr. Seides is no longer with the fund, the remaining partners at Protege Partners appear prepared to make a new wager, this time betting on a sea change in the investment management industry spurred by advancements in artificial intelligence and the maturation of machine learning. In fact, Protege Partners has created a new business, “MOV37”, specifically designed to capitalize on what the partners at Protege Partners see as a highly-probable shift toward “autonomous learning investment strategies”. Autonomous learning investment strategies are technology-driven approaches to investing that leverage computing power and rely on the skill sets of data scientists as opposed to more traditional finance personnel. MOV37 will seek to partner with start-up hedge funds employing these innovative strategies.
Last week at the 4th Annual Ben Graham Value Conference in New York, Protege’s Chief Investment Strategist and MOV37 Partner Michael Oliver Weinberg discussed the new business and shared his views on the investment management industry, past, present, and future. According to Mr. Weinberg, the industry is entering the third wave of its evolution. During the first wave, MBAs and others with similar training and experience had outsized influence and fundamental analysis reigned supreme. The second wave emerged as computing power multiplied and associated costs declined. This enabled quantitative analysis conducted by PhDs to become the ascendant practice. The third wave is currently on the horizon, according to Mr. Weinberg, and it will be characterized by a transition to ALIS and other innovative approaches that utilize machine learning to exploit big data. In the third wave it won’t be PhDs or MBAs in the driver’s seat, but rather computers.
Data is the lifeblood of this third wave and the amount of data available is both substantial and growing. Mr. Weinberg believes data is increasing at a rate equal to Moore’s Law[2], which is to say doubling every two years. IDC, a leading provider of market research on big data, seconds this estimate in their most recent report on the size of the digital universe[3]. Moreover, “unstructured data”, or data that has yet to be processed and stored in an organized database, is expected to grow at a substantially higher rate than data overall. Unstructured data already represents the vast majority of all available data and parsing through this large and rapidly expanding hoard to find ways in which it can be advantageous requires a specific skill set. Demand for data scientists therefore continues to surge. Nevertheless, the amount of unstructured data available is well in excess of what humans can realistically digest. This is where machine learning comes in. If mining data is to yield actionable ideas, it will require a symbiotic relationship between humans & machines. This, according to Mr. Weinberg, is the “ultimate combination”, resulting in superior outcomes to those achievable by people or computers alone. Though possible, a complete eradication of humans from the investment management industry seems improbable. Equilibrium is a more credible end result. Even so, the potential impact that machine learning could have on the investment community is profound.
It is worth noting though that machine learning has been surrounded by hype of this sort for decades, ever since the initial models for artificial neural networks began to emerge during the 1940’s and 1950’s. Neural networks, which simulate how the human brain works, are the foundation of artificial intelligence. They are at the core of Facebook’s facial recognition system and the voice recognition capabilities of Amazon’s “Echo”. But neural networks have had a far less stirring impact on the world of finance. During the 1990’s, many financial institutions attempted to exploit the combination of neural network models and enhanced computing power emerging from the technology boom to bolster their trading strategies. Yet the complexity of the market proved insurmountable and little was ultimately left to show for the effort made and the capital spent. To date, neural networks and machine learning have had an immaterial impact on all but the shortest-term strategies, such as high frequency trading. Other market participants that similarly rely on an informational advantage, but take a longer-term view, are still searching for the optimal way to leverage machine learning.
One market participant that is devoted to finding the solution is Euclidean Technologies. Founded in 2008 by Michael Seckler and John Alberg, two former technology company executives, Euclidean seeks to implement a systematic, technology-driven investment process that is highly reliant on machine learning. Their belief is that technology can be employed to combat the biggest impediment to successful outcomes in investing: human behavioral bias.
Euclidean describes the opportunity thus:
“The fear of bad outcomes – fostered by periodic outcomes that are in fact bad – creates the opportunities Euclidean depends on…
Our goal using machine learning is to endow a systematic process with the intuition – or pattern-recognition skills – of an expert investor. Consider that, lurking in this vast sea of data around public companies and their historical investment outcomes, there may be crucial, non-linear relationships involved in successfully using a company’s operating results to evaluate it as a long-term investment. What might these relationships look like?...
If these hypothetical relationships persisted in the historical record, there is so much about them that linear tools would never see. We do not think great investors are constrained in this way. We believe they evaluate current opportunities in the context of their prior experiences and that those evaluations occur in a non-linear manner. By using machine learning, we aspire to emulate the way an investor builds expertise, but also to inform the learning process with far more examples than any one investor could experience on his own.” (Euclidean Technologies, Q3 2014 Investor Letter)
Beyond utilizing existing machine learning capabilities to automate their process, Euclidian seeks to capitalize on recent innovations such as “deep learning” to exploit the vast and growing reserve of data available to investors. Deep learning is a more advanced form of machine learning that enables computers to go beyond looking for relationships pre-defined by humans, and instead to identify relationships that humans didn’t even know existed.
“Machine learning is, perhaps most simply, the discipline of teaching machines (i.e., computers) how to do things through experience, in a manner that resembles how people learn. …Over the past several years there have been many advances in machine learning, particularly within the field of “deep learning.” Now, more than ever, computers can uncover complex and fruitful patterns that lay hidden in highly complex environments…Perhaps…there are other ways of looking at companies that could be more fruitful? Maybe the tools of deep-learning could open up new areas of analysis that have been previously outside of our grasp?...
The “deepness” in deep learning means that successive layers of a model are able to untangle important relationships in a hierarchical way from data as it is found “in the wild,” with much less pre-processing than has been done in the past. So there is potential to find measures that are more meaningful than what Euclidean relies on today, and also to limit further the potential biases that can impact the learning process when data is processed into forms usable by traditional machine learning tools. …This ability to work with raw information, instead of information that is heavily pre-processed, is in fact one of the important advantages of deep learning. It is part of the reason deep-learning-powered image recognition programs are now performing better than humans in many contexts. (Euclidean Technologies, Q1 2016 Investor Letter)
Image recognition isn’t the only field in which computers have already established themselves as superior to humans. Currently, the world’s foremost players of Checkers, Chess, and Jeopardy! are all computers powered by artificial intelligence. Perhaps most notably, Google’s “AlphaGo” computer recently triumphed over some of the world’s premier players of the complex Chinese game of “Go” - a feat heretofore believed to be impossible. During AlphaGo’s 2016 match against Lee Sedol, one of the most decorated Go players alive, the computer made a decision (on its 37th move) that baffled all the experts. AlphaGo’s lead researcher placed the odds of a human making the same move at 1-in-10,000, yet it ultimately proved to be the difference and resulted in victory. The move served as the genesis for the name of Protege Partners' new fund (MOV37) and it is widely viewed by advocates as an inflection point in the progression of artificial intelligence. The computer analyzed an astronomical amount of data and then made a decision that virtually no human would have made…and it turned out to be right.
Perhaps the complexity of the market will forever remain an insurmountable hurdle for computers, even those with the astounding and expanding capabilities of AlphaGo. Protege Partners is betting otherwise and emerging managers like Euclidean are hard at work, confident that they can find a solution. If their efforts bear fruit it means the success of “move 37” was in fact a precursor to the success of “MOV37”. It also means further concession speeches from Protege Partners will not be forthcoming.
[1] Berkshire Hathaway - 2016 Letter to Shareholders [2] Moore’s Law posits that the number of transistors on an integrated circuit will double approx. every two years. It has held true for 50+ years.[3] IDC: “The Digital Universe of Opportunities: Rich Data and the Increasing Value of the Internet of Things”
Save
