As corporations automate, might the artificial intelligence that runs them one day simply decide to stop distributing profits to human owners? We know the risks automation poses to wages, but what about investment income?
A Path to Executive Automation
Welcome to the rabbit hole of an imaginary entity I call a “perfect profit machine.” It’s a corporation, fully automated all the way up to the artificial intelligence running its highest-level management decisions.
For the sake of this exercise, momentarily set aside your skepticism, and imagine a world where fully automated corporations have actually been achieved. It’s a world of vastly complex decisions processed at lightning speed by artificial intelligence analogous to the executive functioning that organizes human brains. These perfect profit machines would synthesizes unfathomable quantities of data with blinding speed, and a tireless, twenty four-seven, pace with which no human — or group of humans — could compete.
What might actually lead a board of directors to cede corporate control to a machine without processes for vetting decisions with humans? In a future where artificial intelligence beats humans not just in complexity, speed and endurance but in the statistical superiority of its decisions, human interference would simply increase the probability of inferior decision making. As more competitors removed the human throttle from their decision systems, boards would come under tremendous pressure to fully automate, and this new mode of organizational management would spread from company to company and market to market.
Automated Investment Decisions
With no major changes to our economic assumptions between now and the rise of a perfect profit machine, such an entity would operate on the same prime directive that guides corporations today: profit maximization. One of the key challenges this goal would raise is how best to invest operating profits. Undistracted by pride or ego, and with a vastly superior analytical prowess, these systems would wield powerful models for optimizing the way they invest their earnings.
One of the core assumptions in modern U.S. capitalism is that the primary purpose of a corporation is to maximize returns for its shareholders. This is what is known as “shareholder primacy” and it is the code behind the code of today’s publicly-traded corporations. The question is: how would a perfect profit machine come to view distributing its significant portions of its earnings to shareholders as dividends and stock buybacks? Would its artificial intelligence retain shareholder primacy as an overarching goal over time?
This gets to a very interesting question about the evolution of artificial intelligence over time. Does it maintain fidelity to its original programming, or do its instructions mutate and drift over time? In this case, what might happen should shareholder primacy remain as a primary directive, and what might happen if it were to fade away over time?
Automated Shareholder Primacy
To understand the implications of a perfect profit machines guided by shareholder primacy, it helps to look closely at the nature of shareholder relationships with publicly-traded corporations. The most important thing to remember about this relationship is that when shares trade on the stock market, they’re just changing hands with other shareholders. We commonly refer to shareholders as “investors,” based on the belief that their money serves as capital for corporate investments. But even with the exception of Initial Public Offerings and a few other fairly rare situations, the amount of money flowing into corporations is absolutely dwarfed by the amount flowing out through dividends and stock buybacks (which drive up share prices for shareholders).
Over the last 30 years, this transfer of wealth from corporations to shareholders has amounted to an inflation-adjusted $17.474 trillion. Since the year 2000, it’s averaged about 5.5% of the annual U.S. GNP. That is a mind-boggling transfer of wealth. As the vehicle through which this transfer happens, today’s corporations act as powerful engines for extracting value from employees, customers and other stakeholders and concentrating it into the hands of shareholders. This is one of the greatest secrets of our economy, hidden in plain sight within Federal Reserve data.
Now picture this wealth extraction happening at an unimaginably large scale in an economy made up of massive, fully-automated perfect profit machines. The potential explosion in income disparity represents a far more likely threat than the typical terminator concerns often associated with artificial intelligence. It’s likely because it is a direct extension of the reality we experience today. What’s more, it has virtually nothing to do with the technology itself; it’s just the way we choose to use it.
The Temptation of Automated Growth
Let’s now consider a second scenario, where the perfect profit machines shed their shareholder primacy yoke. It might start with the artificial intelligence asking what the corporation actually received in return for the dividends and stock buybacks demanded by shareholders.
Shareholders do create value for companies by making a market for their stock. Through buying and selling, shareholders create liquidity, which matters because it enables a firm to use its stock for things like corporate acquisitions.
We’ve already considered the heavy cost of dividends and stock buybacks, but there is another strategy for generating share liquidity. Shareholders can set aside demands for realized profits when they are able to value a stock on its ability to generate future growth. Jeff Bezos has historically justified Amazon‘s low profits as an investment in future growth.¹ Like other growth stocks, Amazon has tended to reserve that cash to further strengthen its future growth through investments in technological innovation and corporate acquisitions. As those investments bear fruit, their profits in turn fuel future innovation and growth, ad infinitum. This is what is known as a growth stock.
One strategy that a perfect profit machine might use to peacefully subvert shareholder primacy is to simply convince its shareholders that it is a growth stock. This is precisely the strategy we might expect from a perfect profit machine. Its powerful artificial intelligence would excel at optimizing capital investments with a level of complexity and sophistication beyond human understanding. It would master the science of growth, and shareholders would be pleased.
The Risks of Automated Growth
In other words, the perfect profit machine might adopt an aggressive growth stock strategy as a way to free itself from the harness of shareholder primacy. Such a strategy would accelerate the firm’s capital investments in its own automation. As it reaped the competitive advantages of those investments, it would increasingly dominate its markets and look for ways of expanding the reach of its automation into new markets. Competitors would be either driven out of business, or forced to adopt similar strategies in order to survive, thereby accelerating the automation of one market after another. This scenario is far from unrealistic, by the way. Just look at Amazon, which appears to be running a precursor to just such a strategy.
What would be the implications of an arms race between perfect profit machines using this kind of aggressive growth strategy? Nick Bostrom once shared a thought experiment called the “paperclip maximizer:” an artificial intelligence coded with the seemingly innocent goal of making as many paperclips as possible — and it eventually led to a world with lots of paperclips, and no humans.
In short, we need to be careful what we ask for. A world where perfect profit machines enshrine shareholder primacy could lead to horrible stratification of income, but if we’re not careful, the alternative could be much worse.
Automating Stakeholder Stewardship
Rather than leaving things at this dismal conclusion, let us consider a scenario that might initially seem idealistic, but which might actually be quite logical for an artificial intelligence. It shifts the focus from shareholders to stakeholders.
One way to understand stakeholders is “those groups without whose support the organization would cease to exist.” We’ve already talked about the real value that shareholders actually provide to the firm. Might a highly intelligent, perfect profit machine see what today’s CEOs and boards of director fail to see — which is that customers are ultimately far more important than shareholders to the firm’s ongoing success?
Corporations may well automate employees out of work, but human customers are hard to replace. They buy the products and services of the firm, but that’s not all. Customers are no longer the passive consumers once programmed by mass marketing. They actively co-create the value produced by the firm. Facebook would be nothing without its users, and the same is true for hit movies and TV shows, commerce sites like Amazon and Yelp, and the brilliance that is Google Search. Machine learning is already revolutionizing these service, and the data used to train that intelligence simply would not exist without customers participating in these value-creation processes.
The logic of a perfect profit machine might just see through our current fixation on shareholders to the underlying truth that long-term corporate success depends upon the ongoing sustainability of customers, and the communities and ecosystems upon which they rely. This logic builds a self-sustaining feedback loop that strengthens, rather than extracts, value from the stakeholder network that is the corporation.
It’s just a possibility, of course. But if there is a way to mesh economic logic and societal ideals, it might just be through the greater intelligence of a perfect profit machine.
¹Even though Amazon now generates healthy profits, it looks highly unlikely that it would any time soon tap its retained earnings for dividends or stock buybacks.