The beginning of the 20th century, much like the beginning of the 21st, was an era of increasing optimism. The financial panics of 1873 and 1893 were in the past, new technologies like electricity and internal combustion were just gaining traction and Morganization was creating trusts insulated from the ravages of competition.
The financial journalist Alexander Dana Noyes wrote at the time that the market “based its ideas and conduct on the assumption that we were living in a New Era; that old rules and principles and precedents of finance were obsolete; that things could safely be done today which had been dangerous or impossible in the past.”
Yet the old rules still applied and the irrational exuberance led directly to the Panic of 1907. That’s the problem with “new era” thinking. It doesn’t just benefit innovators, it also creates space for marginal players and even outright hucksters. They amplify the boom, but also push risk higher across the system. And when they collapse, they take everyone else with them.
The gravity defying economics of increasing returns
I was working on Wall Street in 1995 when the Netscape IPO hit like a bombshell. It was the first big Internet stock and, although originally priced at $14 per share, it opened at double that amount and quickly zoomed to $75. By the end of the day, it had settled back at $58.25 and, just like that, a tiny company with no profits was worth $2.9 billion.
It seemed crazy, but as economist W. Brian Arthur explained in a 1996 article in Harvard Business Review, certain conditions, such as high upfront investment, negligible marginal costs, and network effects, lead to “winner-take-all markets in which the fastest firm reaps incredible benefits.”
Venture capitalists saw how this could make them rich beyond their wildest dreams and began raising massive amounts of capital to finance the dot-com boom. Calls for deregulation increased, even if it meant greater disruption. Most notably, the Glass-Steagall Act, which was designed to limit risk in the financial system, was repealed in 1999.
The conditions for increasing returns, however, only apply to a narrow swath of businesses, mostly limited to software and electronic gadgets. Nevertheless, entrepreneurs and their investors became convinced that they could apply the Silicon Valley model anywhere, leading to high-profile failures that would soon become all too common.
By 2000, the market peaked, the bubble burst, and the AOL–Time Warner merger became a cautionary tale. While some of the fledgling Internet companies, such as Cisco and Amazon, did turn out well, thousands of others went down in flames. Other, more conventional businesses, such as Enron, WorldCom, and Arthur Andersen, got caught up in the hoopla, became mired in scandal, and went bankrupt.
The false promise of financial engineering
Silicon Valley didn’t invent “new era thinking.” In the 1960s, spurred on by the discovery of an obscure paper written by a long-dead mathematician named Louis Bachelier, economists led by Paul Samuelson drove a revolution in mathematical finance. It would change how financial markets were regulated and managed, with disastrous results.
The operative phrase in Bachelier’s paper, “the mathematical expectation of the speculator is zero,” was as powerful as it was unassuming. It implied that markets could be tamed using statistical techniques developed more than a century earlier. Chicago economist Eugene Fama then built on Bachelier’s work to develop the Efficient Market Hypothesis.
The idea that markets could be rational led to a whole slew of new theories and models, including efficient portfolios, the capital asset pricing model (CAPM), and the Black-Scholes model. That, in turn, gave birth to an entire industry of financial engineering and risk management. Nobel prizes were awarded, fortunes were made, and all seemed well.
Yet even early on there were signs of trouble. In 1963, the mathematician Benoit Mandelbrot published a paper which showed that actual market data exhibited far more volatility than was being predicted by the economists’ models. Other warning signs, such as the collapse of the LTCM hedge fund in 1998, were also ignored.
The idea that mathematical formulas could engineer risk out of the system was just too convenient to give up. It suggested that traders could make almost unlimited profits through sophisticated mathematical engineering. But it was all a mirage, and it came crashing down in the 2008 financial crisis, which nearly collapsed the global economy.
The new era of AI
In February, AI entrepreneur Matt Shumer published a blog post on X titled, Something Big is Happening, that described a stark vision of the future. Amazed by the advances in new AI models’ ability to create code, he announced that AI had crossed a threshold. “I am no longer needed for the actual technical work of my job,” he wrote breathlessly.
Anybody who has used an AI service can see what he means. Machines’ ability to perform human tasks is nothing less than astounding. From writing essays and formatting spreadsheets to helping prepare for a doctor’s appointment or decide what to cook, we are partnering with computers in ways that would have seemed like science fiction just a decade ago.
Much like earlier booms, money has been pouring in. Investment in the technology is set to double this year to $700 billion, most of it going to data centers. Investor Paul Kedrosky estimates that the size of investment has already surpassed that of the dot-com boom and is beginning to approach levels last seen during the railroad frenzy of the 19th century.
Yet even at this early stage, there are worrying signs. A study done by researchers at MIT found that 95% of the companies investing in AI for their employees are getting zero return. A paper by Nobel laureate Daron Acemoglu, which looks at total factor productivity (TFP), a measure which takes capital into account, sees a 0.66% increase over 10 years, translating to a mere 0.064% increase in annual TFP growth. That’s almost nothing.
How can it be that some are so optimistic about AI’s game-changing potential, while at the same time researchers see so little impact? Research by the St. Louis Fed sheds some light. While AI is boosting productivity significantly in some jobs, like those which are related to computers, in other jobs, like in manufacturing and leisure, the impact is minimal.
Dazzled by the world of bits, but still living in the world of atoms
In the 1970s and 80s, business investment in computer technology was increasing by more than 20% per year. Strangely, though, productivity growth had decreased during the same period. Economists found this turn of events so strange that they called it the productivity paradox to underline their confusion.
A paper by researchers at the University of Sheffield sheds some light on what happened. First, there was great disparity between firms in how effectively the technology was deployed. Second, while computers really did transform some tasks, those made up for a relatively small proportion of the overall economy, so the productivity effect was muted.
There’s probably something similar going on now. While artificial intelligence is rapidly transforming the world of bits, we still live largely in the world of atoms. Most of the things we spend money on, like houses, cars, food, and leisure time are, at best, minimally affected by information technology in general and artificial intelligence in particular.
That explains why, despite the emergence of impressive technologies such as the mobile web, big data, and cloud computing over the past twenty years, productivity growth has been mostly depressed over the last 20 years. It also suggests that, much like earlier investment booms, this one will end in a bust.
That’s the problem with “new era” thinking. It’s not that things don’t change. They do. It’s just that everything rarely changes all at once. As I wrote in Mapping Innovation, innovation is never a single event. It’s a process of discovery, engineering and transformation—and the transformation part always takes much longer than anybody thinks it will.