Generative Demand

The Distribution of Artificial Intelligence in "The New Industrial State"

My favorite SORA Demo Video

Demand in the Digital Age

When Sam Altman debuted SORA, OpenAI’s text-to-video product, the announcement and subsequent public fervor felt eerily familiar. In fact, the entire OpenAI product rollout has felt reminiscent of a time in consumer tech not so long ago. It was a time when a professorial Steve Jobs would stand on a minimal stage and unveil the latest iteration of the iPhone, or a tablet, or sometimes an entirely new suite of must-have electronics. I would refresh MacRumors on my first-gen iPhone to see the latest updates and technical specifications. Could the naked eye really detect the marginal improvements in speed and screen resolution? It didn’t really matter; the frontier was coming rapidly into view, and you could be a part of it. 

Artificial Intelligence (AI) adoption is not all that different from the task Jobs and Apple faced a decade ago. It hadn’t previously occurred to me that I needed my phone to be a guitar tuner or a level. Prior to the SORA release, my ability to summon images of dolphins riding bicycles on command had not felt limiting, but the path of technology turns possibility into necessity, into inevitability. 

I am not interested in adjudicating the technology's value, nor am I qualified to predict its path of development. I am concerned with how this technology will be distributed into an economy that has struggled to generate growth in real capital, how intellectual property and computing power available to a small handful of firms can be transformative, and what makes AI different from previous technological step-changes.  

Canadian-American economist, diplomat, and public intellectual, John Kenneth Galbraith provided a model for evaluating these questions two generations ago. Galbraith served as a professor at Harvard University and in various political roles, including U.S. Ambassador to India (1961-1963) and key advisor to Presidents Franklin D. Roosevelt, Harry S. Truman, John F. Kennedy, and Lyndon B. Johnson. Galbraith challenged conventional economic wisdom, emphasizing the importance of countervailing powers and the influence of corporations on consumer behavior.

As untold billions flood into the AI ecosystem—into data centers, chip foundries, and development and model training—what capital—financial, human, political—is being simultaneously deployed into demand generation? AI seems to be on a perpetual PR campaign, but one defined by conflict and contradiction. 

The future with AI is bright but also possibly dangerous. AI will revolutionize the industry, but deployment seems focused on consumer adoption. The main theme of this campaign is that AI is going to change everything, infusing itself into daily life in unimaginable ways. We are to simultaneously believe that AI will be like a utility, yet the firms developing it and its requisite components are the most valuable in the world. The Henry Ford aphorism “If I had asked people what they wanted, they would have said faster horses,” is a familiar defense of speculative innovation, but AI finds itself continually having to prove and promote its qualitative superiority. 

The New Industrial State

Both classical and neoclassical models of economic growth have three basic inputs: Capital, Labor, and Technology. Long-run economic growth ultimately springs from technological progress. The definition of technology, as is common in economics, is impressively broad. Technology includes things as fundamental as fire and row cropping but also microchips, personal computing, and fission. Beyond the scope of economists, it accounts for virtually all of the strange magic that is human progress. 

Capital and labor are somewhat fixed and difficult to alter without considerable turmoil. Technology arises more or less exogenously and improves the productivity of the other inputs. It provides new platforms and infrastructure for industries that didn’t previously exist. Jobs like “app developer,” “podcaster,” or “SEO expert” were not options twenty years ago; high-speed broadband and mobile telephony made them possible, and entrepreneurial creativity took care of the rest.  

The narrative of broadly shared productivity gains was more or less self-evident for much of the 20th century. Technological progress distributed among diffuse small firms in rabid competition led to fantastic economic growth. However, as early as the 1950s, the structure of the US economy began to change. Small firms were superseded by large firms, industrial behemoths that commanded their markets with ever-growing aegis. These firms’ control of both supply and demand was near absolute. Galbraith coined this dynamic as “The New Industrial State,” and he took to task many of the core assumptions of economic growth. 

Central to Galbraith’s theory was that large industrial firms do not risk facing traditional market forces. They do not produce speculatively. Both the time and capital employed in large-scale production would not occur without assurances about the demand for that output. To Galbraith, the New Industrial State was just as preoccupied with generating demand and assuring prices as it was organizing and controlling the means of production.

Unlike fire, row cropping, broadband internet, or mobile telephony, AI is neither a technique nor a platform. It is coveted intellectual property, created at great expense and wholly owned by the firms that develop it. It can’t be built upon like traditional infrastructure; to build an AI product is to consume an AI product. 

The Technostructure

Galbraith referred to the collection of scientists, engineers, marketers, and other specialists as the “Technostructure.” Galbraith's concept of the technostructure refers to the influential group within organizations that possesses specialized knowledge and expertise, enabling them to shape decision-making and wield significant power. 

The Technostructure plays a crucial role in determining the direction and priorities of the organization. The Technostructure's power stems from its ability to control information, set goals, and influence consumer demand through advertising and marketing.

No particular conspiracy is required; no secret cabal needs to meet and coordinate. The modern firm, by siloing talent and controlling resources, accomplishes a dual goal of ensuring its continued ascension while minimizing the potentially disruptive influence of the traditionally capitalist entrepreneur.

Mythmaking around its progenitors has been a cornerstone of the AI PR campaign. The idea that a small menagerie of scientists, visionaries, and entrepreneurs are bringing AI into the world as though it were a divine mission or moral imperative has been sufficiently implanted but the technostructure of AI more closely resembles a mid-century IBM than it does a late-Seventies Apple Computer. 

“The real accomplishment of modern science and technology consists in taking ordinary men, informing them narrowly and deeply, and then, through appropriate organization, arranging to have their knowledge combined with that of other specialized but equally ordinary men. This dispenses with the need for genius. The resulting performance, though less inspiring, is far more predictable.”

These initiatives have been funded and supported by the largest technology companies and investors in the world. Our desire to valorize and mythologize “a couple of guys in their garage” has created a sense that we are all participating in the forging of an unknown frontier. The more likely reality is that AI is merely a component of an intensely planned end-market. 

"Technostructure has emerged as the indispensable planning unit in the modern industrial system. The need for planning arises from the long period of time that elapses during the production process, the high investment in specialized equipment, the inflexible commitment of that equipment to the particular task for which it was designed, and the need for a large organization to which people of diverse skills and functions are committed."

The Perfect Consumer Buys in His Sleep

Galbraith believed that “consumer sovereignty” had meaningfully deteriorated in this New Industrial State; that the economic notion that consumer demands, whether physical needs or psychic desires, were met by entrepreneurs innovating and organizing to meet them, had been upended.  

As society moved beyond subsistence living, consumers required new motivations for consumption. Traditional economics leans on a concept of price-defined utility that is easy to understand in the abstract -- a hungry person places a higher utility on food than leisure -- but harder to parse for post-scarcity affluence. To Galbraith, this sovereignty, where consumer preferences guide production, had been inverted. Galbraith called this the “Dependence Effect” -- that wants ultimately become a function of production. 

“As a society becomes increasingly affluent, wants are increasingly created by the process by which they are satisfied.” 

This helps explain the focus on AI’s consumer-led adoption and how it blurs the line between product and service. AI exponentiates the classic production function, turning its outputs into a funnel for more inputs. AI’s users provide it with the feedstock for its models. The more we use AI, the more robust and desirable its outputs become  -- “wants created by the process by which they are satisfied.” 

But what are those wants? Did consumers and the industry suddenly become consumed with passion for chatbots and derivative image generation? 

“If the individual's wants are to be urgent they must be original with himself. They cannot be urgent if they must be contrived for him.”

Galbraith would hardly believe the current state of what author Shoshanna Zuboff called “Surveillance Capitalism,” the extraction and commodification of personal data for the express purpose of selling. Zuboff writes, 

“The logic of surveillance capitalism begins with unilateral claims to private human experience as free raw material for production and sales. These claims are supported by a radical indifference to what is taken, an indifference that is justified by the logic of the machine at the expense of the interdependent needs of people and society."

This idea is particularly relevant in the context of AI-powered products and services. As consumers interact with AI systems, such as Large Language Models (LLMs), voice assistants, or smart home devices, their preferences and desires are continuously shaped by the outputs of these technologies.

Moreover, AI removes the normal impediments to consumption. Our attention, motivation, or sleep (the last frontier of consumer capitalism) that typically constrain exchange between sentient parties melt away. If adequately supplied with enough information and given sufficient sovereignty on its own, AI can consume by proxy. 

AI is itself a voracious consumer. It requires outsized quantities of both power and water to run through a scarce supply of real estate in the handful of areas even capable of supplying either. Yet at present, flagship AI applications are free to use or can be upgraded for less than the cost of most news subscriptions. An inherent contradiction is that AI is both cheap to use but maddingly difficult to create and enormously expensive to operate.

Subsidized consumption is nothing new, but in earlier iterations, the end game was clear. Uber would operate at a loss to drive out competition from legacy providers, whereas AI is deploying into a greenfield market. A sort of inverse Jevons Paradox is forming, where the technology starts its life as cheap to use, and only after sufficiently embedded is its true cost passed on. Galbraith, 

 “[I]f production creates the wants it seeks to satisfy, or if the wants emerge pari passu with the production, then the urgency of the wants can no longer be used to defend the urgency of the production. Production only fills a void that it has itself created.”

After the last customer service department has been disbanded and after the final studio closes, AI can be profitably deployed. Another form of rent extraction fails to generate the broadly shared economic growth that the classical model promises from new technology. Interactions that were once mediated and limited by human capacity will be carried out more or less continuously and subject to a royalty that covers the upstream requirements of the underlying capital.  

Fewer Demons

“Were it so that a man on arising each morning was assailed by demons which instilled in him a passion sometimes for silk shirts, sometimes for kitchenware, sometimes for chamber pots, and sometimes for orange squash, there would be every reason to applaud the effort to find the goods, however odd, that quenched this flame...”

As Galbraith would say, “The point is so central that it must be pressed.”  While AI may have tremendous utility, it is a product whose production is the justification for its use.

Potential advances in the productivity of low and medium skilled workers, or even their potential obsolescence, would set AI in a more classical economic context. Had our own knowledge work ascended to levels that made these tasks so rote as to be mindless, we could view AI as we now do spreadsheets or e-mail.  Had we found ourselves newly in need of masses of copywriters, data entry clerks or graphic designers, the great movement of capital that brought AI to the fore could be seen in a new light. Instead, we face an aggressively marketed solution applied to glibly defined problems of human capacity. 

This view does not require one to be a techno-pessimist, nor a Luddite. I used a Large Language Model at various points in the composition of this very essay. I am optimistic that in time these technologies will lead to key breakthroughs for challenging problems in health care, energy production and education. 

Galbraith’s view challenges us to see these technologies not as portals, but as mirrors. The path of AI distribution and development relies on our human fallibility and our psychic discomfort in an affluent age. In that analysis, those important breakthroughs likely exist somewhere on the other side of more precise consumer credit ratings, more intrusively targeted advertising and the continued proliferation of various forms of vice.

“...but should it be that his passion was the result of his first having cultivated the demons, and should it also be that his effort to allay it stirred the demons to ever greater and greater effort, there would be a question as to how rational was his solution. Unless restrained by conventional attitudes, he might wonder if the solution lay with more goods or fewer demons.”

John Kenneth Galbraith (1908-2006) Galbraith he told his doctors shortly before his death, “You’re the finest medical team in what was formerly known as the free world.”

---

Investing Implications of Artificial Intelligence and Machine Learning