What Will Happen to the AI Bubble?

 

In an online commentary on The New Yorker website, writer Joshua Rothman tackles the question of the artificial-intelligence (AI) bubble.  On this first week of the new year, that seems like an appropriate question to ask.  It's pretty clear that AI is not going away.  Too many systems have embedded it in their productive processes for that to happen.  But Rothman raises two related questions that only time will answer for sure.

 

The first question is whether the money spent on AI is going to be worth it.  "Worth it" can mean a variety of things.  The most obvious (and frightening, to some) application of AI is direct replacement of workers:  think a roomful of draftsmen replaced by three engineers at computer workstations.  Accountants can most easily justify this way of leveraging AI by showing their managers how much the firm is saving in salaries, offset by whatever the AI system cost.  And assuming the tasks, whatever they were, are being done just as well by AI as they were by people before, the difference is the net savings AI can effect.

 

But as Rothman points out, that approach is both overly simplistic and doesn't reflect how AI is typically being used most effectively.  The most powerful use mode he has found in his own life is to use AI as a mind-augmenting tool.  He gives the example of helping his seven-year-old son write better code.  (I will overlook the implications of what the future will be like with a world full of people who were coding when they were seven.)  ChatGPT helped Rothman find several applications that his son was both able to master, and enjoyed as well. 

 

And in general, the most fruitful way AI is used seems to be as a quasi-intelligent assistant to a human being, not a wholesale replacement.  The problem for businesses is that this sort of employee augmentation is much harder to account for.

 

He points out that if an employee uses AI to become better educated and more capable, that fact does not show up on the firm's balance sheet.  Yet it is a form of capital, capital being broadly defined as anything that enables a firm to be productive.  Rothman cites economist Theodore Schultz as the originator of the term "human capital," which captures the concept that an employee has value for his or her abilities, which can depreciate or be improved just as physical capital such as factory buildings or machinery can be. 

 

In a book I read recently called Redeeming Economics, John D. Mueller points out that modern economic theory simply cannot account for human capital in a logically consistent way.  This constitutes a basic flaw that is still in the process of being remedied.  The usual metrics of economics such as GNP (gross national product) treat investments in human capital such as education and training as consumption, the same as if you took your college tuition and blew it on a vacation to Aruba. 

 

So it's no surprise that businesses are unsure about how to justify spending billions on AI if they can't point to their balance sheets and say, "Here's how we made more money by buying all those AI resources." 

 

Something similar happened with CAD software.  When companies discovered how much more effective their designers were when they began using computer-aided design programs such as AutoCAD, and their competitors began underbidding them as a result, they had to get with the program and spend what it took to keep up. 

 

It's not clear that the results of widespread use of AI will be quite as obvious as that.  Some bubbles are just that:  illusory things that pop and leave no significant remnants.  Rothman cites a rather cynical writer named Cory Doctorow who believes the AI bubble will pop soon, leaving scrap data servers and unhappy accountants all over the world. 

 

But other bubbles turn out to be merely the youthful exuberance of an industry that was just getting established.  A good example of that kind of bubble was the automotive industry in the years 1910 to 1925.  There were literally dozens of automakers that popped up like mushrooms after a rain.  Most of them failed in a few years, but that didn't take us back to riding horses. 

 

Both Rothman and I suspect that the AI boom, or bubble, will be more like what happened with automobiles and CAD software.  The feverish pace of expansion will slow down, because anything that can't go on forever has to stop sometime.  But the long-term future of AI depends on the answer to Rothman's second question:  how good will AI get?

 

It's clearly not equal in any general sense to human intelligence today.  As Rothman puts it, AI assistants are "disembodied, forgetful, unnatural, and sometimes glaringly stupid."  These characteristics may simply be the defects that further research will iron out in ways that aren't obvious.

 

While I'm not in the market for a job right now, I nevertheless receive lists of possible jobs from my LinkedIn subscription.  A surprising number of them lately have been what I'd call "AI checking" jobs:  companies seeking a subject-matter expert to make queries of AI systems and critique the results.  Clearly, the purpose of that is to fix problems that show up so the mistakes aren't made the next time.

 

It's entirely possible that some negative world event will trigger an AI panic and rush to the exits.  But even if the short-term spending on AI does crash, we still have come a long way in the last five years, and that progress isn't going to go away.  As Rothman says, AI is a weird field to try and make forecasts for, because it involves human-like capabilities that are not well defined, let alone well understood.  My guess is that things will slow down, but it's unlikely that humanity will abandon AI altogether, unless some terrifying doomsday-sci-fi tragedy involving it scares us away.  And that hasn't happened so far.

 

Sources:  Joshua Rothman's article "Is A. I. Actually a Bubble?" appeared on Dec. 12 on The New Yorker website at https://www.newyorker.com/culture/open-questions/is-ai-actually-a-bubble?.  I also referred to John D. Mueller's Redeeming Economics (ISI Books, 2010), pp. 84-86. 

 

Clifford Furnas and the Clouded Crystal Ball

In 1936, during the depths of the Great Depression, a professor of physical chemistry at Yale named Clifford C. Furnas published a book in which he tried to anticipate the next great advances in science and engineering during the following century.  His book was inspired by a visit he made to the Chicago World's Fair in 1933, otherwise known as the "Century of Progress Exposition," which marked the 100-year anniversary of the founding of Chicago.  A lot of the technical exhibits that were designed to show how the world of tomorrow would be better than the depression of today didn't work properly, and so he went home and surveyed the state of science, engineering, and technology and made his best guesses as to how things would be by 2033, appropriately entitling it The Next Hundred Years.

My interest isn't so much in the accuracy of his technical predictions as in his expectations for what the trend of automation would yield for the economy and the working life of the average citizen.  It was already obvious by 1933 that a lot of jobs formerly done partly or wholly by hand up to then would be performed by machines or even robots in the future.  But what Furnas missed, along with nearly every other prognosticator up to the end of World War II, was the rise of the electronic computer, computer networking, and the growth in Internet-based economic activity.  And without the computer, modern robotics would be impossible, because without digital control systems (now including artificial intelligence), a robot can't do anything much more than act as a power-assist to a human being.

What we're talking about is the rise in what economists call productivity:  the economic output of a nation divided by the number of hours worked.  One person using a small lathe and a few hand tools can build a watch in maybe a few dozen hours, depending on what they start with.  But one person at the controls of an otherwise fully automated watch factory can make hundreds or thousands of watches per hour.  And Furnas was right in his prediction that advances in automation would (a) greatly increase the productivity of the average worker, and (b) render obsolete entire classes of jobs that previously employed millions of people. 

Where he went wrong was his prediction about what the result of these changes would be.

In Furnas's view, the average man (he barely discussed women at all), when faced with a choice of working 40 or 50 hours a week for ever-increasing pay, or else getting paid the same wages for less and less work, would choose to work less and get paid the same amount for it.  Consequently, the great challenge he foresaw for the future was to find things for people to do with all their spare time, now that their jobs could be done in as little as one or two hours a day.  He summarized the difficulty thus:  "Our problem will be to keep the citizenry on even keel while they have a wealth of time on their hands, for certainly a society steeped in mere idleness will soon lose its moral fiber, its material possessions and its reasons for existence." 

Why didn't things turn out that way?  Why isn't the U. S. a peaceful country full of debating societies, painting groups, and volunteer choirs, instead of harboring an increasingly divided populace in which some better-educated folks live a life of relative freedom and interesting work, while most people without advanced degrees work longer and longer hours in uncertain dead-end jobs (sometimes two or three jobs at once) and feel they can barely get by?  And don't forget the growing class of working-age men who have simply resigned from the workforce altogether and spend their days playing video games and in other forms of, in Furnas's words, "mere idleness."

A complete answer to these questions would require a book, or several books by a group of experts with talents that I lack.  But in my 300 words or so remaining, I'll hazard a few guesses.

One answer will sound paradoxical:  the rise in the standard of living.  The phrase "keeping up with the Joneses" captures some of this idea.  For Furnas's vision of the leisure class to come to pass, it wouldn't do for just a few people to choose shorter working hours over more pay—most of the country would have to do it.  And in the hyper-competitive international economic arena, a country in which most of its working people work only two hours a day would lag behind countries where 40 or 50 hours a week was the norm. 

Another answer is that people are, frankly, greedy.  And greed, at least of the mildly acquisitive type, is the engine that fuels advertising and consumer economies such as in the U. S. and most other industrialized nations these days.  There are a few people who choose to live on next to nothing and cut themselves off from the grid, but most of us regard them as eccentrics at best and dangerous at worst. 

A third factor is what I call "building-code creep."  If you attempted to build a house today in the way a modestly-priced house was built in 1930, you would be violating nearly every building code in the book.  Where's the third wire for grounding the outlets?  Where's your insulation, air conditioning, smoke alarms?  What's all this lead paint doing here?  That gas water heater has no automatic flameout-protection valve.  In thousands of  ways that have made life safer and more convenient, we have changed the rules of material life so that it costs a great deal more to live simply than it used to.  In certain rural parts of the country, most if not all of these things can be skipped, but at the price of living dangerously.

For a variety of reasons, we seem to be entering a period in which increasing numbers of people in the U. S. choose to live without jobs.  But most of them don't seem to be happy about it, and I think Furnas was on to something when he expressed concern about the deteriorating moral fiber of a nation where idleness becomes a way of life for many people.  The key, if there is one, lies in the phrase "reasons for existence," but that is a topic for another blog.

Sources:  Clifford C. Furnas's The Next Hundred Years was published in 1936 by Reynal & Hitchcock, New York.  The quotation about keeping citizens on an even keel is from p. 367.  I previously referred to this book in my blog on Sept. 23, 2013, "Engineers and Technological Unemployment:  What Are People For?"