The Productivity Myth: Behind The Economic Pitch for A.I.
It could cause large scale job loss! But create jobs too! OpenAI is honing a complicated and potentially contradictory narrative about A.I.
It’s been a big week for artificial intelligence in Washington D.C. with the release of the Trump White House’s A.I. Plan and all the attendant hype around and hobnobbing between the tech elite and the political establishment alongside it.
As a follow up to the recently failed proposal from Congressional Republicans to outlaw most A.I. regulation, the White House is moving forward with its own pro-industry push, including a report Wednesday that called for A.I. to be freed from “onerous regulation” and for the U.S. to speed its growth through the streamlined permitting of data centers, manufacturing facilities and other infrastructure needed to power the energy-intensive tech.
Amidst this backdrop, OpenAI CEO Sam Altman swung into town to warn about job loss and fraud in a speech hosted by the Federal Reserve, while company underscored the big economic benefits it says A.I. will bring. Confusing and contradictory? A bit. But let’s do a closer read.
OpenAI used the week for some big announcements, chief among them the opening of a D.C. headquarters and a newly honed pitch for A.I., the engine behind its biggest product, ChatGPT.
For this week’s newsletter, I wanted to unpack that pitch, because the narrative around A.I. seems as important as its impact right now. Since it’s impossible to know the future that A.I. will bring about, and even the technology itself is not perfectly understood by experts, the story around it is an especially important factor in how it is viewed, implemented and ultimately regulated — or not.
And of course, this a labor story at its heart.
A New Story
The story about OpenAI’s new messaging surfaced this week in Axios, in a brief written by Mike Allen, whose writing has served as an unofficial main stage for the Washington establishment for years.1 For those whose brains have been poisoned by D.C. politics long enough to read the signs, but not so much to fall prey to them, his file can be revealing, serving as a sort of x-ray image of the consensus manufacturing process in D.C. — who will be making which case and why, what talking points they will be using, and how the political PR battle that is being fought daily will unfold across the week.
In this case, Allen was given information from someone in Altman’s camp about a new set of points the CEO and OpenAI were planning to use in their lobbying efforts for A.I. On the heels of warnings that Altman and others have made about losing the A.I. arms race to China, a pro-industry argument last spotted as one of the main justifications behind the attempt to prevent AI regulation, Altman and OpenAI were now going to trot out an economic pitch, Allen reported.
They were going to argue A.I. needs to be used widely in order for its economic benefits to be both fully realized and fairly spread. The OpenAI source called this a promise to keep AI “democratic.”2
To bolster this push, the company also released a report this week from its in-house economist, Ronnie Chatterji, that made two substantial points. One: that adoption of A.I. by everyday people, businesses and professionals has been rapid and widespread. The paper cited internal data, which was also given to Axios in advance, that ChatGPT users send more than 2.5 billion prompts every day across the world; more than 330 million of those are in the U.S.
The second point the report made is key: that A.I. is being used by people and firms to increase productivity and free up workers for more critical tasks. To back this up, Chatterji cited a handful of recent studies about how various professionals — teachers, lawyers, government workers, consultants and others — have been able to save time by outsourcing work to A.I., freeing themselves up for more important tasks.3
Axios’ anonymous source laid out the underlying ethos of the pitch here: "OpenAI sees AI fundamentally as a productivity-driving technology…The big question isn't if it will grow the economic pie — it's who gets how much of a slice ... It's not about stopping disruption, but putting it into people's hands so they have the opportunity to benefit."
There’s a lot here to untangle. The OpenAI rep does nod to concerns that the economic benefits of A.I., like those of previous tech booms, could accrue in unequal ways. Ok, check. But then comes a more mixed message: the “disruption” should not be stopped — disruption that may wipe out of half of all entry-level white collar jobs or 30 percent of all jobs by 2030 — but it must be widespread enough to bring benefits.
Disruption is no longer shiny linguistic currency it used to be for the tech industry; it’s been shaded by darker tones as that upheaval has gotten more intense. So the company seems to be acknowledging negative downstream effects, and saying they will be balanced out by a strong benefit: increased productivity.
The Productivity-Pay Gap
To weigh this argument, it’s important to look at how productivity has worked in the U.S. economy for the last 50 years.
Economists have long preached the gospel of increasing productivity, arguing that more worker output, through innovation, efficiency, and better infrastructure, education and other means, leads to economic gains which are then shared broadly. This means better wages, conditions for workers, higher standards, and obviously more production – in short, a better and more prosperous economy.
This equation bore out for many years. If you look at old charts, the upward slant of productivity was well-aligned with the hockey stick arrow of higher wages in the U.S.
But something that started in the 1960s began to worsen by the 1980s. Productivity increases decoupled from average wages — rising, while average wages stagnated.
The Economic Policy Institute, the left-leaning thinktank in D.C. that produced the chart above, gives a few reasons behind this decoupling: failures to adequately raise the minimum wage,4 decades-old labor laws that are ill-fit to the modern business arena, lower tax rates on top income brackets, and deregulatory pushes, including in anti-trust, finance and a wide-swath of every day industries, that have hurt workers.
Where did the windfall from increased productivity go? Into increased salaries for higher-level corporate and professional employers, and larger company profits that manifest as returns to shareholders, investors, and owners.
We are still early in the A.I. growth curve. We’re continuing to learn about its uses and economic effects. And this is not to say or conclude that it can’t be a boon.5 But it’s important to note that increased productivity is not an inevitable or perhaps even a likely salve to the problem of large scale job loss, worsening inequality, or other economic pitfalls on its own. In fact, for people who believe that inequality undergirds much of the societal rancor that has, you might say, slightly upended things, it would seem to risk making things even worse, without proper guardrails.
Here are a few other pieces we’re reading this week:
Do y’all use Reddit? I’m a big fan these days, using it to source pasta techniques, vintage waffle irons, road trip recommendations, advanced stain removal methods, wellness advice, and the never-ending sports chatter that I use to quiet the howling void in my soul more often than I’d like to admit. Apparently I’m not alone — this piece in New York mag charts Reddit’s exploding popularity and looks at it as a sort of last vestige of the early web: open, weird, and blissfully human. But now maybe under threat? Le sigh.
I liked this profile of former top White House Russia expert, later Trump admin skeptic Fiona Hill in the New York Times.
Eric Jaye, a Bay Area-based political consultant who I recently spoke to for another project, made a succinct argument here that the focus on Biden’s age, in the post-mortem push (or non-push) to understand the Democrats’ election failure, in Jake Tapper and Alex Thompson’s “Original Sin,” and elsewhere, is the result of Harris staffers and Democratic consultants who don’t want to face accountability for losing a winnable campaign.
This scapegoating succeeds, in part, because of the incestuous relationship between the Democratic consultants and the national political press. There has historically been a bias among journalists for the senior campaign consultants who can help make a story with a quote, or drop a timely exclusive. Now the operatives can market their services by filling the chairs in the non-stop cable news pundit blasts. The two sides feed, and need, each other more than ever.
This interview with Julie K. Brown, the reporter at the Miami Herald whose investigation of Jeffrey Epstein’s exceedingly light sentence in 2008 put his case and criminal activities back on the map, is a must-read or listen for anyone who is plugged into this story.
Back on the A.I. front, this 17-minute long mini-documentary, about a tiny town outside of Memphis being faced with potentially illegal levels of air pollution from a massive new data center that Elon Musk and xAI rapidly assembled recently in what the video implies is a violation of the Clean Air Act, is astounding and worth a watch as a shiver-inducing parable about the way things work these days.
See you all next week.
Allen’s most recent column, as of this writing, begins:
President Trump, in terms of raw accomplishments, crushed his first six months in historic ways. Massive tax cuts. Record-low border crossings. Surging tariff revenue. Stunning air strikes in Iran. Modest inflation. Yet poll after poll suggests most Americans aren't impressed. In fact, they seem tired of all the winning….This paradox is unfolding in what could be the very best chapter of his presidency…”
This appears to be used as a synonym for “widely used,” or available to all. Are social media services “democratic,” because of their availability?
My colleague William Fitzgerald, who had a long career at Google previously, notes that Google began issuing reports that highlighted its role in boosting the economy around the time regulators began to take a closer look at the company more than a dozen years ago.
As one example: the current federal minimum has been stuck in place since 2009.
Partially as an experiment, I often use A.I. to do a quick copy edit of my pieces, checking for syntax errors, typos or other grammatical mistakes. It does an ok job; not nearly as good as the Professional Human Copy Editors (PHCEs) I used to work with in the Wood Pulp, Ink And Words Business, but any additional All-Natural Intelligence (ANI) labor, beyond your two Hard Reset co-authors, is out of the budget at the moment. But props where props are due, I guess: the artificial machine offered up the word “currency,” above, as an unsolicited recommendation during its copy edit that fit in quite nicely. It also complemented my piece. Gee thanks A.I, you really know the way to this every writer’s heart.