ChatGPT Is About To Revolutionize The Economy. We Need To Decide What That Looks Like

An artificial intelligence gold rush has begun over the past few months to mine the expected business opportunities from generative AI models like ChatGPT, whether it is based on hallucinatory beliefs or not.

In an effort to understand the sensational text-generating bot that OpenAI published last November, app developers, venture-backed startups, and some of the biggest corporations in the world are all frantically trying to understand it.

Around the world, corner offices can virtually be heard shouting, “What is our ChatGPT play? How can we profit from this?

However, while businesses and executives clearly see an opportunity to profit, it is much less clear how the technology will affect employees and the economy as a whole. ChatGPT and other recently released generative AI models hold the promise of automating a variety of tasks that were previously thought to be solely within the realm of human creativity and reasoning, from writing to creating graphics to summarizing and analyzing data. Despite their limitations, chief among which is their propensity for making stuff up, ChatGPT and other recently released generative AI models are not without their own limitations. Because of this, economists are uncertain of how jobs and total productivity may be impacted.

Despite the amazing developments in AI and other digital tools over the past ten years, it is disheartening to see how little they have contributed to increased wealth and broad-based economic growth. Few investors and business owners have become extremely wealthy, but the majority of people haven’t profited. Some people have even lost their employment due to automation.

Since around 2005, productivity growth in the US and the majority of advanced economies has been poor, which is how nations become richer and more prosperous. (the UK is a particular basket case). The reality that the economy is not growing much has caused many people’s wages to stagnate.

In that period, there has been some productivity growth, but it has primarily been concentrated in a small number of industries, like information services, and in a small number of US cities, like San Jose, San Francisco, Seattle, and Boston.

Will ChatGPT exacerbate the already alarming wealth and revenue disparities in the US and many other nations? Or would it be useful? Could it really give work the much-needed boost?

Both the recently released DALL-E 2 and ChatGPT, which can write in a manner akin to a person, use large language models that have been trained on vast amounts of data. The same can be said for competitors like Bard from Google and Claude from Anthropic. These so-called foundational models, like OpenAI’s GPT-3.5, which ChatGPT is built on, or Google’s rival LaMDA, which powers Bard, have advanced quickly in recent years.

The amount of parameters—variables in the models that are altered—is drastically increasing, and they are being trained on an increasing amount of data. GPT-4, the most recent version of OpenAI, was published earlier this month. OpenAI won’t specify the precise size difference, but based on GPT-3’s 175 billion parameters and comparison to GPT-2, one can make an educated guess.

But for many users, everything shifted after ChatGPT was released toward the end of last year. It is compelling in its ability to quickly produce human-like text, including recipes, exercise schedules, and—perhaps most surprisingly—computer code. It is also extremely simple to use. The user-friendly chat model, which is less abstract and more practical than the impressive but frequently esoteric advancements that have been brewing in academia and a small number of high-tech companies over the last few years, is unmistakable proof that the AI revolution has real potential for many non-experts, including a growing number of entrepreneurs and businesspeople.

Venture capitalists and other investors are pouring billions of dollars into generative AI-based businesses, and the number of applications and services powered by big language models is expanding daily.

Among the major players, Microsoft has reportedly invested $10 billion in OpenAI and its ChatGPT in the hopes that the technology will breathe new life into its ailing Bing search engine and give its Office products new powers. Early in March, Salesforce revealed that it would integrate a ChatGPT app into its well-known Slack platform and also unveiled a $250 million fund for generative AI startup investments. The list continues on and on, from GM to Coca-Cola. Everybody plays ChatGPT.

Google, meanwhile, declared that Gmail, Docs, and some of its other well-known products would use its new creative AI tools.

There are still no glaring killing applications. Additionally, economists claim that as companies race to find new applications for the technology, a rare opportunity has arisen to reconsider how to maximize the advantages of the new generation of AI.

“We’re having this conversation now because this technology is available to handle. You can now interact with it without knowing any programming languages. Many people can begin to envision how this affects their workflow and job prospects, according to Katya Klinova, the Partnership on AI’s director of research on AI, labor, and the economy.

“Who will profit, is the question? Klinova, who is working on a report outlining the potential employment impacts of generative AI and offering suggestions for utilizing it to increase shared prosperity, asks, “And who will be left behind?

The upbeat perspective is that it will prove to be a useful instrument for many workers, enhancing their skills and knowledge while boosting the economy as a whole. The pessimistic one: Businesses will merely use it to eliminate well-paying jobs that once appeared to be automation-proof and that demand creativity and logical thinking; a small number of high-tech companies and tech elites will get even richer, but it will have little impact on overall economic development.

Helping the least skilled

ChatGPT’s effect on the workplace is a real issue, not just a speculative one.

According to the most current analysis, 80% of the US workforce may be impacted in some way by large language models like GPT, according to Tyna Eloundou, Sam Manning, Pamela Mishkin, and Daniel Rock of the University of Pennsylvania. They also predicted that GPT-4 and other anticipated software tools would have a significant impact on 19% of jobs, with at least 50% of those positions being “exposed” to AI models. They indicate that higher-paying jobs would be most impacted, which is in contrast to what we observed in earlier waves of automation. Writers, web and digital designers, finance quantitative analysts, and—just in case you were considering a career change—blockchain engineers are some of the professions that are most at risk.

David Autor, a labor economist at MIT and a recognized authority on the effects of technology on the workforce, asserts that “there is no doubt that [generative AI] is going to be used—it’s not just a novelty.” It’s already being used by law companies, and that’s just one instance. It makes a variety of chores that can be automated possible.

Author has spent years demonstrating how highly sophisticated digital technologies have eliminated numerous well-paying manufacturing and routine clerical jobs. He claims that ChatGPT and other generative AI examples have altered the computation, though.

In the past, AI had automated some office duties, but those were the repetitive, sequential tasks that could be programmed for a machine. Now it is capable of performing jobs that we previously considered creative, like writing and creating graphics. Anyone paying attention will quickly realize that generative AI opens the door to the computerization of numerous chores that are currently thought to be difficult to automate.

The concern is that businesses will replace relatively well-paying white-collar jobs with this new form of automation, sending those workers off to lower-paying service employment while the few who are best able to exploit the new technology reap all the benefits. As Autor points out, there are plenty of jobs in the US.

In this case, tech-savvy employees and businesses might adopt the AI tools rapidly and boost their productivity to the point where they rule their industries and workplaces. Those who had less technical knowledge and ability to begin with would fall further behind.

However, Autor also foresees a more advantageous outcome: generative AI might assist many people in acquiring the necessary skills to compete with those who possess greater schooling and experience.

Such a result might be feasible, according to one of the first thorough studies on ChatGPT’s productivity effect.

Shakked Noy and Whitney Zhang, two MIT graduate students in economics, conducted an experiment with hundreds of college-educated professionals working in fields like marketing and HR, asking half of them to use ChatGPT everyday while the other half did not. The fact that ChatGPT increased productivity overall isn’t all that shocking, but here’s what’s really intriguing: the AI tool benefited the least qualified and accomplished employees the most, narrowing the performance difference between workers. In other words, the bad writers improved significantly while the excellent writers only slightly increased their speed.

The early results indicate that ChatGPT and other generative AIs may be able to “upskill,” in the language of economists, job-seekers. Many skilled workers are “lying fallow” as a result of losing their positions in industry and offices over the past few decades, according to Autor. Our workforce may be revitalized if generative AI can be used as a useful tool to broaden their knowledge and give them the specialized skills necessary in fields like teaching or health care, where there are plenty of employment opportunities.

The situation that prevails will depend on how carefully we consider how we want to use the technology.

“I don’t believe we should accept it because technology is out there and we have to get used to it. It can be used and developed in a variety of ways because it is still being made, according to Autor. It’s difficult to overstate how crucial it is to build for its intended purpose.

Simply stated, we are at a turning point where either less skilled workers will be able to do more of what is currently considered knowledge work, or the most talented knowledge workers will dramatically increase their current advantages over everyone else. Which result we end up with largely relies on how employers use tools like ChatGPT. The more optimistic choice, however, is easily within our grasp.

Beyond human-like

Brynjolfsson contended that the quest for human-like abilities has resulted in technologies that merely replace people with machines, driving down wages and escalating wealth and income disparity. He called it “the single biggest explanation” for the growing wealth disparity.

A year later, he claims ChatGPT, with its human-sounding outputs, “is like the poster child for what I warned about,” having “turbocharged” the conversation about how new technologies can be used to enhance rather than replace people’s abilities.

Brynjolfsson, the director of the Stanford Digital Economy Lab, is usually a techno-optimist when it comes to artificial intelligence, despite his concerns that AI developers will continue to blindly outdo each other in mimicking human-like capabilities in their creations. He anticipated a rise in productivity from AI and other digital technologies two years ago, and today he is optimistic about the effects of the new AI models.

Brynjolfsson’s optimism is largely based on his belief that organizations could gain a lot from utilizing generative AI, such as ChatGPT, to broaden their product offerings and boost employee productivity. It’s a fantastic instrument for creativity. It excels at assisting you in trying out new things. Not just doing the same job less expensively, according to Brynjolfsson. He asserts that it will be crucial as long as businesses and developers can “avoid the mentality of thinking that humans aren’t needed.”

He estimates that within a decade, generative AI could boost US economic development by trillions of dollars. According to him, knowledge workers and information workers make up the bulk of the workforce in our country. And it’s difficult to imagine any category of computer workers that won’t be at least somewhat impacted.

If and when that increase in output occurs is a matter of economic conjecture. Maybe all we need to do is wait.

The MIT economist Robert Solow, who was awarded the Nobel Prize in economics that year for illuminating how innovation fuels economic growth, is credited with saying, “You can see the computer age everywhere except in the productivity statistics,” in 1987. The effects, especially from advancements in semiconductors, didn’t start to appear in productivity data until much later, in the mid- and late-1990s, as businesses began to figure out how to benefit from ever-cheaper computational power and associated advancements in software.

Could AI lead to a similar outcome? It depends on whether we can work out how to use the most recent technology to transform businesses as we did in the earlier computer age, according to economist Avi Goldfarb at the University of Toronto.

He claims that businesses have so far only used AI to perform certain duties a little bit better: “It will increase efficiency—it might incrementally increase productivity—but ultimately, the net benefits are going to be small. Considering that all you’re doing is the same thing slightly better. The technology, however, “doesn’t just enable us to do what we’ve always done a little bit better or a little bit cheaper,” he adds. It might enable us to develop new procedures to provide consumers with value.

The jury is still out on whether or when generative AI will lead to that. According to Goldfarb, “the big productivity boost will come once we understand what good writing at scale allows industries to do differently, or—in the context of Dall-E—what graphic design at scale allows us to do differently.” But I don’t know if that will be in a week, a year, or in ten years.

Power struggle

When the new crop of large language models, like ChatGPT, became available, economist Anton Korinek of the University of Virginia and a fellow at the Brookings Institution did what many of us did: he started experimenting with them to see how they might benefit his research. In a paper published in February, he meticulously evaluated their performance, noting how well they managed 25 “use cases,” including text editing and brainstorming (both very helpful), coding (pretty good with some assistance), and math. (not great).

According to Korinek, ChatGPT did erroneously describe one of the core ideas of economics: “It screwed up really badly.” But given the advantages, the simple error was soon overlooked. As a cognitive worker, “I can tell you that it makes me more productive,” he claims. “There is no doubt in my mind that using a language model increases my productivity.”

He evaluated the performance of GPT-4 using the same 25 questions that he had recorded in February, and it fared much better. There were fewer cases of making up information, and it performed significantly better on the math homework, according to Korinek.

According to Korinek, an increase in economic output could occur much faster than in previous technological revolutions because ChatGPT and other AI bots automate cognitive work rather than physical tasks that call for investments in infrastructure and equipment. By the end of the year—certainly by 2024—I believe productivity may experience a larger boost.

Additionally, he claims that in the long run, the ability of AI models to increase the productivity of researchers like himself has the potential to propel technical advancement.

Research in the physical disciplines is already revealing this potential of large language models. Machine learning is a specialist field for Berend Smit, who directs a chemical engineering facility at EPFL in Lausanne, Switzerland. Smit asked one of his graduate students, Kevin Maik Jablonka, to show that GPT-3 is actually useless for the complex machine-learning studies his group conducts to predict the properties of compounds last year after he presented some intriguing findings using it.

He was a total failure, Smit quips.

It turns out that the model works as well as sophisticated machine-learning tools especially created for chemistry at answering fundamental questions about things like the solubility of a compound or its reactivity after being fine-tuned for a few minutes with a few pertinent examples. It only requires a compound’s name to forecast a variety of properties based on the structure.

Large language models, like those used in other fields of work, have the potential to increase the knowledge and skills of non-experts—in this case, scientists who are unfamiliar with sophisticated machine-learning techniques. Jablonka claims that because it is as easy as conducting a literature search, “it could bring machine learning to the masses of chemists.”

These impressive—and unexpected—results are only a tantalizing glimpse of how potent the new AI technologies could be across a variety of creative endeavors, including scientific finding, and how startlingly simple they are to employ. But this also raises some important issues.

Who will establish the vision for how these tools should be developed and used as the potential impact of generative AI on the economy and employment grows more imminent? Who will be in charge of this incredible science in the future?

One worry, according to Diane Coyle, an economist at Cambridge University in the UK, is the likelihood that the same major corporations that control a significant portion of the digital world will come to dominate large language models. She notes that Google and Meta are supplying their own extensive language models in addition to OpenAI, and that entry into the market is made difficult due to the high computational costs needed to operate the software.

According to Coyle, the concern is that these businesses share comparable “advertising-driven business models.” “So obviously, if you don’t have different types of people with different kinds of incentives, you get a certain uniformity of thought.”

A publicly funded international research group for generative AI, modeled after CERN, the Geneva-based intergovernmental European nuclear research body where the World Wide Web was invented in 1989, is one solution, according to Coyle, who admits that there are no quick fixes. It would have the substantial computing capacity required to run the models and the specialized scientific knowledge needed to advance the technology.

“Bring some diversity to the incentives that the creators of the models face when they’re producing them,” claims Coyle of such an endeavor outside of Big Tech.

Despite the fact that it is still unclear which public policies would ensure that large language models best serve the public interest, according to Coyle, it is becoming obvious that the decisions about how we use the technology cannot be left to a small number of market-dominant businesses.

We have plenty of instances from history that show how crucial government-funded research can be in creating technologies that lead to widespread prosperity. The US Department of Defense sponsored ARPANET, which invented methods for multiple computers to interact with one another, and this publicly funded effort gave birth to the internet long before the web was created at CERN.

The MIT economists Daron Acemoglu and Simon Johnson offer a compelling tour through the history of technological progress and its mixed record in generating broad prosperity in their book Power and Progress: Our 1000-Year Struggle Over Technology & Prosperity. Their argument is that it’s crucial to purposefully direct technological advancements in ways that offer widespread advantages rather than just enhancing the wealth of the elite.

The US economy underwent rapid technological change from the post-World War II era until the early 1970s. During this time, the majority of employees saw increases in pay while income inequality significantly declined. According to Acemoglu and Johnson, this is because technological advancements were used to create new jobs and tasks, and political and social constraints made sure that employees shared the rewards of these advancements with their employers more fairly than they do today.

They claim that, in comparison, the Midwest, which is considered to be “the industrial heartland of the American economy,” has experienced a “prolonged regional decline” due to the more recent rapid adoption of manufacturing robots.

The book, which will be published in May, is especially pertinent for comprehending the potential effects of the current rapid advancements in AI and how future decisions regarding how best to apply them will impact each and every one of us. Acemoglu claimed in a recent interview that they were working on the book when GPT-3 first came out. And we anticipated ChatGPT, he continues half-jokingly.

According to Acemoglu, those who are developing AI “are going in the wrong path.He claims that the AI’s complete architecture is currently in automation mode. But nothing about generative AI or AI generally should compel us in this path. It is the business strategies and outlook of those working for Microsoft, OpenAI, and the venture capital industry.

If you think we can change the course of technology, then it begs the question, “Who are “we”?” Furthermore, Acemoglu and Johnson are at their most controversial here. In their article, they state: “Society and its strong gatekeepers need to stop being mesmerized by tech billionaires and their goals. One need not be an expert in artificial intelligence to have an opinion on how these technologies will shape our civilization in the future.

Much of the credit for introducing the public to the new AI sensation belongs to the developers of ChatGPT and the businesspeople engaged in its commercialization, particularly Sam Altman, CEO of OpenAI. It has a lot of promise. But that doesn’t mean we have to agree with their ideas about where technology should go and how it should be applied.

Their narrative states that the ultimate goal is artificial general intelligence, which, if everything proceeds according to plan, will result in enormous economic prosperity and abundance. The vision has lately received extensive promotion from people like Altman, who has long advocated for a universal basic income (UBI) to support those of us who aren’t technocrats. It sounds enticing to some people. Free money and no labor! Sweet!

The narrative’s fundamental assumptions—that AI is inevitably going to destroy jobs and that most of us are just along for the (free?) ride—are what concern me the most. This perspective barely acknowledges the chance that generative AI might help people outside of the tech-savvy elites unlock their talents and brains, resulting in a boom in creativity and productivity. The concept of utilizing technology to spread prosperity among the working populace by increasing human capabilities and expertise is rarely discussed.

Acemoglu and Johnson state in their article: “We are going toward greater inequality not unavoidably but due to bad decisions about who has power in society and the direction of technology… In reality, UBI fully subscribes to the business and tech elite’s belief that they are the wise, capable individuals who should generously fund the rest of society.

Acemoglu and Johnson outline a number of strategies for creating “a more balanced technology portfolio,” ranging from tax changes and other governmental initiatives that might promote the development of AI that is more worker-friendly to changes that might wean academia off Big Tech’s funding of computer science research and business schools.

These changes are “a tall order,” the economists concede, and a social movement to refocus technological change is “not just around the corner.”

The good news is that we do have control over how we use ChatGPT and other sophisticated language models. Businesses and individual users will be able to decide how they want to use ChatGPT as a plethora of apps based on the technology rush to market. Businesses can use ChatGPT to give employees more abilities—or to simply eliminate jobs and trim costs.

Another encouraging development is that open-source generative AI initiatives are starting to gain traction and may eventually overthrow Big Tech’s control of the models. Notably, over a thousand experts from different countries worked together last year to develop Bloom, a sizable language model that can produce text in languages like French, Spanish, and Arabic. And if Coyle and others are correct, greater government support for AI research could influence the direction of future innovations.

Brynjolfsson, of Stanford, declines to express optimism about the outcome. He clearly enjoys the technology of today, though. If we use technology properly, he asserts, “we can have one of the best decades ever.” However, it’s not unavoidable.

ChatGPT Is About To Revolutionize The Economy. We Need To Decide What That Looks Like

An artificial intelligence gold rush has begun over the past few months to mine the expected business opportunities from generative AI models like ChatGPT, whether it is based on hallucinatory beliefs or not.

Download The Radiant App To Start Watching!