Artificial Intelligence May Not Take Your Job After All

 

 

While artificial intelligence will automate certain tasks and job functions, the human capacity for creativity, critical thinking, and emotional intelligence ensures that many jobs will remain relevant and irreplaceable.

 With the rise of programs converting text to convincing videos and the imminent integration of OpenAI's bots into consumer products, the boundaries of artificial intelligence’s (AI's) impact on daily life are expanding rapidly. The age of generative artificial intelligence (AI) has well and truly come. OpenAI's robots, which use technology called large-language-model (LLM), got things going in November. Now, hardly a day goes by without a new development that blows our minds. A fake Drake and The Weeknd sang on a song made by AI that recently shook the music business. Content made by programs that turn text into video is pretty compelling. Soon, customer services like Expedia, Instacart, and OpenTable will connect to OpenAI's bots. This will let people order food or book a vacation by typing text into a box. A presentation that was recently leaked, which is said to have been made by a Google worker, shows that the tech giant is worried about how easy it is for competitors to make progress. There will likely be a lot more to come.

The advancement of AI brings up a lot of important questions. The question that is easiest to answer is often one of the most important: What exactly does this imply for the state of the economy? A great number of people have really high hopes. According to recent research conducted by the bank Goldman Sachs, widespread AI adoption could eventually drive a 7% or almost $7trn increase in annual global GDP over a ten-year period. This growth would take place over the course of the next decade. Academic studies suggest a gain of three percentage points in yearly labor-productivity growth in enterprises that implement the technology. This would represent a significant increase in salaries that would be compounded over a number of years. According to a research by Tom Davidson of Open Philanthropy, an organization that makes grants, there is a greater than 10% possibility that explosive growth could occur at some point in this century. Explosive growth in this regard is defined as increases  in global output of more than 30% a year. A handful of economists have entertained the idea that global earnings could one day reach an endless level.

However, financial markets seem to indicate results that are rather less dramatic. In the past year, the performance of the share prices of companies active in AI has lagged behind that of the worldwide average, despite the fact that these share prices have increased during the past few months. The interest rates are still another piece of the puzzle. If people believed that technological advancement was going to make everyone affluent tomorrow, interest rates would increase since there would be less of a need to save money. According to studies conducted by Basil Halperin of the Massachusetts Institute of Technology (MIT) and colleagues, inflation-adjusted rates and eventual GDP growth are closely connected with one another. However, ever since the AI hype began in November 2022, long-term interest rates have been on the decline. They continue to be quite low when compared to previous standards. In plain terms, financial markets are not expecting a high probability of AI-induced growth acceleration, on at least a 30-to-50-year time horizon.

To figure out which group of researchers in this area is right, it helps to look at how technology has changed in the past. This is good news for businesses. For one thing, it's hard to say that a single new tool has ever changed the economy a lot, either for the better or for the worse. Even the industrial revolution of the late 1700s, which many people at the time think was caused by the invention of the spinning jenny, was really caused by a number of different things happening at the same time. These included increasing use of coal, firmer property rights, the emergence of a scientific ethos, and much more besides.

 In the 1960s, Robert Fogel wrote about the railroads in the United States. This work would later win him the Nobel Prize in economics. Many people thought that the railroad changed America's future, changing it from a farming society to an industrial powerhouse. Fogel found that it had a very small effect because it replaced technologies, like canals, that would have done just as well. If railways had never been invented, America would have hit the same level of income per person on March 31, 1890, instead of January 1, 1890.

No one can say for sure where AI, a system that is inherently unpredictable, will take humans. Runaway growth is not impossible, and neither is a standstill in technology. But you can still think about what might happen. And, at least so far, it looks like Fogel's railways will be a good guide. Think about monopolies, labor markets, and productivity.

A new technology" can sometimes give vast economic power to a small group of people. John D. Rockefeller won at processing oil, and Henry Ford won at making cars. Jeff Bezos and Mark Zuckerberg are two of the most powerful people in the world today. Many experts think that the AI business will soon make a lot of money. In a recent paper, analysts at Goldman predict that, in the best case, generative AI could add about $430 billion to annual global enterprise-software sales. Their math is based on the idea that each of the world's 1.1 billion office workers will buy a few AI gizmos for a total of $400.

Any company would be happy to get some of this money. But $430 billion doesn't change the big picture of the economy. Assume that all of the income goes into profits, which isn't likely, and that all of these profits are made in the U.S., which is a little more likely. Even so, the ratio of the country's business profits before taxes to its GDP would rise from 12% to 14%. That is a lot higher than the long-term average, but not any higher than in the second quarter of 2021.

One organization—possibly Open AI—might receive these proceeds. When an industry has significant fixed costs or when it is difficult to switch to competitors, monopolies frequently develop. Rockefeller's oil was the only option available to customers, and they were unable to make their own. There are certain monopolistic traits in generative AI. One of Open AI's chatbots, GPT-4, reportedly cost more than $100 million to train, which is money that not many businesses have on hand. A lot of proprietary knowledge about the data used to train the models also exists, in addition to user feedback.

However, there are slim chances of one business dominating the entire AI sector. It is more likely that a small number of large businesses will compete with one another, as is the case in the grocery, search engine, and airline industries. Since all AI products employ the same models, none are truly unique. This makes switching from one to another easier for a customer. Additionally, the models' computing power is comparatively generic. Since a lot of the code and helpful hints are publicly available online, amateurs can create their own models—often with startlingly impressive outcomes.

In general, generative AI doesn't seem to have any systemic moats at the moment. A similar conclusion is drawn in the current leak that is allegedly from Google: The barrier to entry for training and experimentation has decreased from the complete output of a major research organization to one individual, one evening, and a powerful laptop. A few generative-AI companies now have a market value of over $1 billion. The largest corporation to date to benefit the most from the new AI era is not even an AI firm. Data center income is growing at Nvidia, a computing firm that powers AI models.

 Doomsday or comedy show?

Despite the possibility that generative AI won't produce a new class of robber barons, for many people, that is still a cold comfort. They are more worried about their own financial future, particularly whether their job will disappear. There are many dreadful forecasts. Numerous sources predict that at least 10% of the work tasks performed by at least 80% of the US workforce could be impacted by the introduction of LLMS. According to studies from several periodicals, the professions of legal services, accounting, and travel agencies  are among those most likely to experience disruption.

Economists have made doomsday forecasts in the past.  Many people in the 2000s questioned how outsourcing might affect jobs in developed nations. In a widely read report published in 2013, two Oxford University researchers predicted that technology would eliminate 47% of American jobs over the next few years. Others argued that even in the absence of widespread unemployment, there would be a hollowing out of the labor market, where rewarding, well-paying occupations would vanish and be replaced with mindless, poorly paid jobs.

What eventually happened caught everyone off guard. The average unemployment rate in the rich countries has roughly decreased by half during the past ten years. The proportion of people in working age who are employed is at an all-time high. The least unemployment is seen in nations with strong levels of automation and robots, such Japan, Singapore, and South Korea. According to a recent analysis by the American Bureau of Labor Statistics, jobs identified as at risk from new technologies did not exhibit any general tendency toward notably rapid job loss in recent years. Evidence of hollowing out, if any, is mixed. The 2010s saw an increase in work satisfaction metrics. The poorest Americans have had faster pay growth than the richest Americans throughout the majority of the last ten years.

It might be different this time. Chegg, a company that provides homework help, recently saw its share price drop by 50% when it acknowledged ChatGPT was having an impact on their new customer growth rate. Big IT corporation IBM's CEO stated that the company anticipates pausing hiring for positions that AI may eventually replace. But are they early warning signals of an impending tsunami? Maybe not.

Imagine a scenario in which more than 50% of the tasks covered by AI are automated. Or consider that job losses are based on the percentage of jobs that are mechanized across the entire economy. According to estimates provided by various media, this would in either case result in a net loss of  about 15% or more of American jobs. Some people might relocate to a sector like hospitality where there is a dearth of workers. However, a significant increase in the unemployment rate would undoubtedly follow—possibly matching the 15% briefly attained in America during the worst of the COVID-19 pandemic in 2020.

However, this scenario is unlikely to occur because history indicates that job destruction occurs far more gradually. To replace human operators, the automated telephone switching system was created in 1892. The Bell System didn't set up its first fully automated office until 1921. The number of American telephone operators increased even after this landmark, reaching a peak of about 350,000 in the middle of the 20th century. Despite the invention of automation nine decades prior, the profession did not (largely) go out of existence until the 1980s. In fewer than 90 years, AI will completely transform the work market: LLMS are simple to use, and many experts are surprised by how quickly people have adapted ChatGPT into their daily lives. However, the same factors that led to the delayed uptake of technology in previous organizations still hold true now.

There are  plenty of other published evidence whose argument in this area focuses on regulation. The pace of technical advancement is typically pitifully slow in sectors of the economy where the government is heavily involved, such as education and healthcare. Absence of competition dampens motivation to progress. Additionally, governments may have objectives for public policy that conflict with increased effectiveness, such as maximizing employment levels. Additionally, unionization is more prevalent in these industries, and unions are effective at preventing job losses.

There are numerous examples of these. Although there is technology to partially or completely replace train drivers, for instance, they are nevertheless paid close to twice the national median on London's publicly run Underground network. Passengers  must repeatedly fill out paper forms for government organizations with their personal information. Real-life cops are still employed in San Francisco, the epicenter of the AI boom, to control traffic during rush hour.

 When Bots Go Wild

 The majority of AI-threatened jobs are in heavily regulated sectors. According to published evidence, 14 of the top 20 occupations most vulnerable to AI are teaching positions, with foreign-language instructors near the top and geographers in a slightly stronger position. However, only the most courageous government would replace teachers with AI. Think of the headlines. The same holds true for police officers and anti-crime AI. The fact that Italy has already temporarily blocked ChatGPT over privacy concerns, and France, Germany, and Ireland are reportedly mulling the option, demonstrates how concerned governments are about the job-destroying effects of artificial intelligence.

Perhaps, over time, governments will permit the replacement of some jobs. However, the delay will allow the economy to continue doing what it does best: creating new categories of jobs as others are eliminated. By reducing production costs, new technology can increase demand for products and services, thereby boosting jobs that are difficult to automate. The conclusion of a report published in 2020 by MIT's David Autor and colleagues was startling. Approximately 60% of American jobs did not exist in 1940.  The year 2000 saw the addition of fingernail technician to the census. Five years ago, solar photovoltaic electrician was introduced. The AI economy will likely create occupations that are currently unimaginable.

Small labor-market effects are expected to have a small impact on productivity—the third element. Electricity was first used in companies and homes in America near the end of the 19th  century. However, there was no production boom until the end of World War I. In the 1970s, the personal computer was invented. This time, the productivity surge came faster—but it still felt slow at the moment. An economist named Robert Solow famously said in 1987 that the computer age was everywhere except the productivity statistics.

The world is still waiting for a spike in production due to recent advances. Smartphones have been widely used for a decade, billions of people have access to ultrafast internet, and many professionals now work from home when it is convenient for them. Official polls reveal that well over a tenth of American employees now work for companies that use AI of some kind, while unofficial studies show even higher numbers. However, global productivity growth remains subdued.

AI has the potential to significantly increase productivity in various industries. Erik Brynjolfsson of Stanford University and colleagues investigated customer-service representatives in their research paper. The availability of an AI tool increases the number of issues handled per hour by 14% on average. Researchers may also become more efficient as a result of GPT-X, which may provide them with a limitless number of almost-free research helpers. Others expect that AI will eliminate administrative inefficiencies in health care, hence lowering costs.

However, there are many things that AI cannot do. Construction and farming, which account for around 20% of the rich-world GDP, are two examples. A LLM is useless to someone harvesting asparagus. It may be useful to a plumber who is fixing a leaky tap: a widget could recognize the tap, diagnose the problem, and recommend remedies. However, the plumber must still perform the physical labor. So it's difficult to imagine blue-collar labor being significantly more productive in a few years. The same is true for industries where human-to-human contact is an essential component of the service, such as hospitality and medical care.

AI is also powerless to address the single most significant impediment to rich-world productivity growth: misfiring planning systems. People cannot live and work where they are most efficient when city sizes are limited and housing expenses are high. No matter how many amazing new ideas your society has, they are rendered functionally useless if they are not built in a timely manner. It is up to governments to silence not in my back yard (NIMBYS) protesters. Technology is both here and there. The same is true for electricity, where permitting and infrastructure maintain costs unbearably high.

It equally possible that the AI economy will become less productive. Consider some current technology. Smartphones provide for immediate communication, but they may also be a source of distraction. Email keeps you linked 24 hours a day, seven days a week, which might make it difficult to concentrate. According to the best available research, the more time spent on email each day, the poorer perceived productivity.  Some managers now worry that working from home, which was once viewed as a productivity booster, provides too many individuals with an excuse to slack off.

The use of generative AI may reduce productivity. What if AI can generate entertainment that is exactly customized to your every need, for example? Furthermore, few people have considered the ramifications of a machine that can generate massive amounts of text in an instant. For a NIMBY facing a planning application, GPT-4 is a blessing. He can write a well-written 1,000-page complaint in five minutes. Someone must then respond to it. Spam emails will become more difficult to identify. Fraud cases could skyrocket. Banks will have to spend more money on avoiding attacks and compensating victims.

Battle of Words

In a world with a lot of AI, there will be more lawyers. In the 1970s, you could make a multimillion-dollar deal on 15 pages because retyping was such a pain. AI will let us cover the 1,000 most likely edge cases in the first draft, and then the parties will fight about it for weeks. In America, a good rule of thumb is that you shouldn't sue for damages unless you want $250,000 or more in pay, since it costs that much to go to court. Now, the costs of going to court could drop to almost nothing. Teachers and writers will have to check everything they read to make sure it wasn't written by an AI. OpenAI has put out a tool that makes this possible. So, it is giving the world a way to deal with a problem that its own technology has caused.

AI could change the world in ways that we can't even think of right now. But this is not the same as flipping the economy on its head. Fogel wrote that his case wasn't meant to disprove the idea that the railroad was a key part of America's growth in the 19th century. Instead, it was meant to show that the evidence for this idea is not nearly as strong as is usually thought. When studying generative AI in the middle of the 21st century, a future Nobel Prize winner may well come to the same result.

The fear of job displacement due to AI automation is prevalent, but history shows that job destruction happens slowly. Regulatory barriers and heavily regulated sectors, like education and healthcare, may delay the replacement of certain jobs. Additionally, new technology creates new types of jobs that cannot be imagined today.

 

 

Notes

 

Autor, D., Mindell, D., & Reynolds, E. (2020). The Work of the Future: Building Better Jobs in an Age of Intelligent Machines. Retrieved from MIT: https://workofthefuture.mit.edu/wp-content/uploads/2021/01/2020-Final-Report4.pdf

Brynjolfsson, E., Li , D., & Raymond, L. R. (2023). Generative AI at Work. National Bureau of Economic Analysis - Working Paper. doi:10.3386/w31161

Cohan, P. (2023, May 25). Nvidia Stock Soars As CEO Jensen Huang Cites Generative AI Edge. Retrieved from Forbes: https://www.forbes.com/sites/petercohan/2023/05/25/nvidia-stock-soars-as-ceo-jensen-huang-cites-generative-ai-edge/?sh=98562a66eec9

Davidson, T. (2021, June 25). Could Advanced AI Drive Explosive Economic Growth? Retrieved from Open Philanthropy : https://www.openphilanthropy.org/research/could-advanced-ai-drive-explosive-economic-growth/

Donadel, A. (2023, March 24). College Professors Face the Highest Exposure to AI Tools, Study Finds. Retrieved from University Business: https://universitybusiness.com/college-professors-face-the-highest-exposure-to-ai-tools-study-finds/

Fogel, R. W. (1964). Railroads and American Economic Growth: Essays in Econometric History. Baltimore, MD: The Johns Hopkins University Press .

Halperin, B., Chow, T., & Mazlis, J. Z. (2023, January 10). AGI and the EMH: Markets are Not Expecting Aligned or Unaligned AI in the Next 30 Years. Retrieved from Effective Altruism Forum: https://forum.effectivealtruism.org/posts/8c7LycgtkypkgYjZx/agi-and-the-emh-markets-are-not-expecting-aligned-or

Krishnan, M., Mischke, J., & Remes, J. (2018, June 4). Is the Solow Paradox back? Retrieved from McKinsey Quarterly: https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/is-the-solow-paradox-back

McCallum, S. (2023, April 1). ChatGPT Banned in Italy Over Privacy Concerns. Retrieved from BBC News: https://www.bbc.com/news/technology-65139406

McKendrick, J. (2023, May 26). Most Jobs Soon To Be ‘Influenced’ By Artificial Intelligence, Research Out Of OpenAI And University Of Pennsylvania Suggests. Retrieved from Forbes: https://www.forbes.com/sites/joemckendrick/2023/03/26/most-jobs-soon-to-be-influenced-by-artificial-intelligence-research-out-of-openai-and-university-of-pennsylvania-suggests/?sh=693029ab73c7

Moy, C. (2023, January 30). Spinning Jenny: Who Invented the Spinning Jenny? Retrieved from The Economic Historian: https://economic-historian.com/2022/07/spinning-jenny/

National Geographic. (2023). Industrial Revolution and Technology. Retrieved from https://education.nationalgeographic.org/resource/industrial-revolution-and-technology/

Osborne, M., & Frey, C. B. (2013). Automation and the Future of Work – Understanding the Numbers. Technological Forecasting and Social Change, 114, 254-280.

Patel, D., & Ahmad, A. (2023, May 4). Google "We Have No Moat, And Neither Does OpenAI". Retrieved from Semi-Analysis: https://www.semianalysis.com/p/google-we-have-no-moat-and-neither

Prakash, P. (2023, May 2). Chegg’s Shares Tumbled Nearly 50% After the Edtech Company Said Its Customers are Using ChatGPT Instead of Paying for its Study Tools. Retrieved from Forbes: https://fortune.com/2023/05/02/chegg-shares-tumble-students-fleeing-chatgpt-a-i/

Price, D. A. (2019). Econ Focus: Goodbye, Operator. Retrieved from https://www.richmondfed.org/publications/research/econ_focus/2019/q4/economic_history

The Economist. (2023, May 7). Beyond the Hype: Your Job is (Probably) Safe from Artificial Intelligence. Retrieved from https://www.economist.com/finance-and-economics/2023/05/07/your-job-is-probably-safe-from-artificial-intelligence

U.S. Bureau of Labor Statistics. (2022, July). Growth Trends for Selected Occupations Considered at Risk From Automation. Retrieved from Monthly Labor Review: https://www.bls.gov/opub/mlr/2022/article/growth-trends-for-selected-occupations-considered-at-risk-from-automation.htm

Veltman, C. (2023, April 21). When You Realize Your Favorite New Song Was Written and Performed by ... AI. Retrieved from NPR: https://www.npr.org/2023/04/21/1171032649/ai-music-heart-on-my-sleeve-drake-the-weeknd#:~:text=Press-,'Heart%20on%20my%20Sleeve'%20uses%20AI%20to%20simulate%20Drake%20and,raises%20legal%20and%20ethical%20questions.