The AI Octopus

ERIC POSNER

Source Link

With long-gestating antitrust cases against Google, Apple, and Amazon coming to fruition, many observers think that 2024 could be a turning point for Big Tech. Yet even as authorities press ahead with this litigation, they risk being blindsided by the rise of artificial intelligence, which is likely to reinforce Big Tech’s dominance of the economy.

The recent firing and rehiring of OpenAI CEO Sam Altman was interpreted as a conflict between cautious board members who worried about the risks of AI and enthusiasts like Altman. But the real significance of this episode was what it revealed about OpenAI’s relationship with Microsoft, the largest investor in OpenAI’s commercial operations. While OpenAI’s nonprofit structure nominally means that only its board controls it, the board was forced to rehire Altman after Microsoft expressed misgivings that helped instigate an employee revolt.

Microsoft is not just an investor in OpenAI; it is also a competitor. Both companies develop and sell AI products, and Microsoft had stopped short of acquiring OpenAI to avoid antitrust problems. But if Microsoft controls or partly controls OpenAI, the two companies may have an illegal collusive relationship. That is why the US Federal Trade Commission and the United Kingdom’s competition authority are both investigating the matter.

The OpenAI-Microsoft relationship is only a small part of a rapidly growing AI oligopoly. As a recent paper by the law professors Tejas Narechania of the University of California, Berkeley, and Ganesh Sitaraman of Vanderbilt University documents, market power is rife all along the AI supply chain. The monopolist Nvidia manufactures most of the chips needed for AI development. Amazon, Google, and Microsoft dominate cloud computing, which is essential for storing the data on which AI models are trained.

Since these companies and Meta (Facebook/Instagram’s parent company) are among the only ones that can collect and store such data, they are also the ones developing – and profiting from – the most important AI models and applications. While Microsoft has a lock on ChatGPT, along with its own proprietary AI apps, Google has Bard and, alongside Amazon, is investing billions in Anthropic (the developer of Claude).

Virtually all the major tech companies and their executives are connected through institutions and professional networks. These include the start-up incubator Y Combinator (where Altman served as president before moving to OpenAI); joint research projects (such as an AI-related partnership that includes Google, Facebook, Amazon, and Microsoft); corporate boards; and social relationships. OpenAI’s own nonprofit board includes, or has included, people with connections to other companies that are also developing AI products. One of its founding members, for example, was Elon Musk, who quit several years ago.

This web of close connections creates significant opportunities for collusion (which is illegal) or coordination (which is legal but still bad for consumers). But with public attention focused largely on individual Big Tech firms’ abuses of monopoly power, people have overlooked the many ways that tech companies can collude with each other to extend their market power.

These behemoths have a long history of such behavior. In 2010, several of them settled a case, brought by the Department of Justice, in which they were accused of agreeing not to hire away one another’s software engineers. In emails, Apple’s Steve Jobs scolded Google’s Eric Schmidt for allowing recruiters to hire Apple employees, and Schmidt ordered a subordinate to fire a recruiter “verbally” so as to “avoid a paper trail.”

In an ongoing case challenging Google’s dominance of search, the court heard evidence that Google paid Apple for default status on the iPhone, with the possible goal of keeping Apple out of the search market that Google monopolizes. In another case against Google for monopolizing digital advertising, the DOJ alleges that Google paid off Facebook to avoid a challenge to its stranglehold on that market.

Years ago, Apple was caught orchestrating a collusive scheme among book publishers. And when OpenAI’s employees threatened to leave for Microsoft after Altman’s ouster, they were effectively trying to sell OpenAI to Microsoft. It was a kind of collusion among employees to accomplish a merger that would have violated the spirit, and possibly the letter, of antitrust law.

The cozy relationships among tech executives are reminiscent of the Gilded Age “money trust” of key banks and financial institutions that both supplied capital to the era’s industrial giants and colluded with them and one another. The money trust’s extraordinary power led to antitrust legislation (in 1890 and 1914), regulation (including the establishment of the US Federal Reserve in 1913), and eventually laws that broke up banks, restricted their involvement in owning companies, and limited their operations (in the 1930s). Unlike an oil company or a railroad, banks are uniquely positioned to drive consolidation across the economy, because they can use their financial leverage over virtually every company in diverse economic sectors to control their behavior, including by pushing for mergers.

Big Tech firms now resemble banks in their influence across the economy – but at a supercharged level. Through their access to data, they know more about consumer and business behavior, and exert more control over it, than banks ever did. They supply vital inputs to businesses across the economy, as well as products and services to almost all consumers. No bank has ever had such reach.

It is little wonder that tech companies are also displacing financial institutions at the economy’s commanding heights. As a rueful financier commented in the Financial Times, Big Tech has consistently shouldered aside financial institutions in the race to buy AI companies. Not only are the six largest US-based companies (by market capitalization) tech firms, but the smallest among them (Meta) is almost twice the size of JPMorgan. The top seven tech companies now account for 30% of the entire S&P 500; even in its 1920s heyday of market domination, the banking system accounted for only 16-19%.

Cartoonists from the early era of antitrust depicted the financial connections among the banks and real-economy monopolists as octopus tentacles, which encircled and squeezed politicians and consumers alike. If AI lives up to its promise and becomes the lifeblood of every sector of the economy – as an input in every industry from law to manufacturing – we can expect a future of economic concentration and corporate political power that dwarfs anything that came before.