Doing Capitalism in the Innovation Economy's Author Bill Janeway Offers Some Context
Q: How did you first get interested in economics and finance?
A: I am the son of a practicing political economist, Eliot Janeway, who focused on the triangular interactions between Washington and Wall Street and the real economy out there on the Main Streets of the country and the world. So I was immersed in this world of practice and theory from an early age. As an undergraduate at Princeton in the Woodrow Wilson School, my more formal education in financial economics and related policy studies accelerated. I wrote my undergraduate thesis on the economic policies of the Hoover Administration in the US and their failure in the face of the financial crisis that precipitated the Great Depression. And at Cambridge University, the subject of my doctoral dissertation was “The Economic Policies of the Labour Government of 1929”: you may detect a certain pattern here. I then spent what I call my 35-year sabbatical from the academy in the world of entrepreneurial finance at the technological frontier. When I returned to Cambridge University just in time for the Global Financial Crisis of 2008, my focus shifted to the dynamics of investment at that frontier, where progress is made by trial and error and funding must come from sources that are not constrained by the requirement of immediate, visible economic value. So the central subject of my book, Doing Capitalism in the Innovation Economy, is the complementary role of mission-driven states and productive financial bubbles in financing transformational technological innovation.
Q: And then what factors led to your involvement in investment banking?
A: I became an investment banker at the start of the 1970s largely by accident. I had decided not to pursue an academic career after receiving my Ph.D. from Cambridge and, on the other hand, two summers working in Washington during the politically destructive escalation of the Vietnam War had vaccinated me against what is known as “Potomac Fever.” As described in chapter 1 of my book, I “stumbled” into an extraordinary investment firm, F. Eberstadt & Co., founded by a forgotten titan of mid-twentieth century America, Ferdinand Eberstadt. The firm’s activities in investment research, investment management, and investment banking were concentrated on the science-based industries: chemicals, pharmaceuticals, electronics, and the nascent computer industry. It was at Eberstadt that I became immersed in computing and information technology generally. It was also at Eberstadt that I made the transition from investment banking agent to venture capital principal.
Q: How did you first begin to hear about the use of A.I and machine learning in finance?
A: In the second chapter of my book, I recount the engagement of myself and my colleagues in the Hype Cycle of AI that occurred from the late 1970s to the mid-1980s. This reflected the fundamentally mistaken belief that capturing decision rules from human experts and translating them into software programs would enable computers to exercise human judgment. In fact, it enabled computers to perform the rigid practices of rule-bound apprentices. Expert systems have not disappeared, of course: they are embedded in virtually every software program whose execution requires looking up an appropriate function to execute under explicitly pre-defined conditions. In this, so-called expert systems were a grandiose extension of what by the 1980s were already broadly deployed in the financial sector, as computerized trading systems began to dominate the market. Wall Street’s adoption of such systems was the result of the “Back Office Crisis” at the end of the 1960s, when the archaic, manual trading and settlement systems broke down under the impact of increased volume generated by the rise of institutional investors.
Q: When did finance become computerized?
A: One of the lessons to be learned from the failure of expert systems was that the inputs to human judgement often take the form of pattern recognition. Now, the stock market since time immemorial has been an arena where investors and their advisors have sought to establish predictable outcomes of observable patterns in trading behavior. Pre-computer “chartists” developed and deployed a whole array of visible signals of patterns such as “head and shoulders” and the “double bottoms.” So, Wall Street was bound to be a happy hunting ground for the inventors of new, more sophisticated methodologies for pattern recognition, especially when computers enabled such methodologies to trigger buy and sell orders increasingly closer to real time. I well recall the revolution in Wall Street that represented the response of the agents – the investment banks – to the demands of their dominant customers – the investing institutions – for best execution of their orders at the lowest net price. As I put it in my book, they “spawned a set of bigger, smarter, tougher competitors who made unimaginably more money as principals than they ever could have as agents” (p, 22).
Q: How do you assess the impact of AI and machine learning on finance?
A: The net of this is that to my mind, it is necessary to separate the different sorts of techniques for data analysis from the marketing brand of AI. Extracting reliable information from data is only the first step, problematic as that itself can be. From that information, extracting actionable meaning from the information is a process different in kind: for “meaning” is always and inevitably context-dependent. There are two sorts of domains where the current generation of machine learning techniques have generally proved to be quite effective. The first is where there is an independent “objective correlative,” such as a physical object in the case of image recognition. The second is where there are fixed, exogenously determined rules of a game whose state is always transparently visible to all players. However, in the world in which humans interact with humans, “context” and therefore “meaning” is always and inevitably contestable and to become stable requires negotiation to an agreed, if generally transient, state. This is why I share the view of such authorities as Rodney Brooks, emeritus Director of MIT’s Computer Science and AI Lab, that current machine learning techniques do not represent the path to “Artificial General Intelligence.”
Q: What potential do you see with the use of AI in finance? What fears do you have about the potential use of AI?
Q: Final Thoughts?
Written by Ramsay Bader, Edited by Alexander Fleiss
About William H. Janeway:
William H. Janeway is a Special Limited Partner of Warburg Pincus. He joined Warburg Pincus in 1988 and was responsible for building the information technology investment practice. Previously, he was executive vice president and director at Eberstadt Fleming. Dr. Janeway is a director of Magnet Systems and O'Reilly Media. He is an Affiliated Member of the Faculty of Economics at Cambridge University.
Dr. Janeway is a co-founder and member of the board of governors of the Institute for New Economic Thinking. He is a member of the board of directors of the Social Science Research Council and of the Advisory Board of the Princeton Bendheim Center for Finance. He is a member of the management committee of the Cambridge-INET Institute, University of Cambridge and a Member of the Board of Managers of the Cambridge Endowment for Research in Finance (CERF). He is the author of Doing Capitalism in the Innovation Economy: Reconfiguring the Three-Player Game between Markets, Speculators, and the State, the substantially revised and extended new edition of the book initially published by Cambridge University Press in November 2012.
Dr. Janeway received his doctorate in economics from Cambridge University where he was a Marshall Scholar. He was valedictorian of the class of 1965 at Princeton University.