The promise and peril of artificial intelligence as a general purpose technology

And unlike early general-purpose technologies, AI appears to operate at the top end of the skills distribution. Image: Shutterstock

Unless you’ve been living under a rock, most of us would have, over the course of the past half-year, encountered some astounding uses of artificial intelligence (AI) technology. This could have been the shocking ability of ChatGPT to generate long essays responding to virtually any topic. Or the uncanny skill of Syntheses in producing deepfake videos of strangers and celebrities alike. Or the fearful ease that tools such as Github Copilot, Tableau, and Consensus have been able to replicate all manners of skilled work, ranging from programming to data analysis to producing entire research papers.

This has forced many white-collar workers to confront a suddenly-uncertain future. Until now, the specter of technological progress had always come for the blue-collar job. Centuries ago, the mechanized loom threw thousands of hand weavers out onto the streets, while the spread of the direct-dial telephone after the Second World War made the services of manual switchboard operators redundant. More recently, the personal computer and smartphone replaced the stenographer and secretary, both low-end, white-collar functions.

The response of the displaced workforce has, hitherto, always been to climb the skills and value-added ladder. In place of bricklayers and ditch diggers, we trained construction workers who knew how to handle pneumatic drills, operate excavators, and put together prefabricated buildings. We invested in human capital that would be comfortable with algebra and calculus and statistics, sophisticated medical equipment, and advanced manufacturing techniques. We got our children to learn to code, so that they could manipulate the machines that dispatched their parents from their jobs.

But AI as a general-purpose technology—of which large-language models and generative AIs such as ChatGPT are but the latest iterations—is unlike previous waves. The diffusion of technology has always accelerated over time. Electrification, for instance, took about a half century before becoming ubiquitous, while computers and the Internet took at least a generation. Smartphones have spread more rapidly, finding their way into the hands of even those in rural Africa and Asia in a little more than a decade. But even compared to companies at the vanguard of Web 2.0—such as Facebook and Twitter, which took between a year to two to gather a million users—ChatGPT has become pervasive even more rapidly, having gotten there in a mere 5 days.

Also read: How AI in travel and hospitality is enhancing experiences and optimising operations

And unlike early general-purpose technologies, AI appears to operate at the top end of the skills distribution. While the computer scientists that design algorithms for AI and other machine learning models will likely still find their skills relevant in an AI-driven economy, a great many others will find themselves replaced. Five years ago, we had already seen examples of how routine mental work—such as cleaning spreadsheets, editing text, and playing games—could be easily automated away. But AI can now outperform even white-collar workers engaged in creative, cognitively-demanding, nonroutine tasks, such as designing websites, performing radiographic diagnoses, and summarizing legal briefs. Pair AI with advances in robotics that have become ever-better at fine and gross motor functions—from piecing together microchips to navigating military obstacle courses to driving cars and trucks—and it becomes difficult to imagine what sort of jobs aren’t immune to the threat from this newest wave of general-purpose technology.

Of course, some of this fear is exaggerated. If history is any guide, the leaps made in the earlier stages after the introduction of general-purpose technologies peter out over time. After an initial period of disruption, completely new applications—once inconceivable to contemporaries—would find their way into common use. Producers of Gone with the Wind and the Wizard of Oz could scarcely have imagined that, one day, their industry would become one where actors would perform in front of a green screen, alongside computer-generated co-stars, all digitally delivered and enjoyed predominantly at home.

And along with these novel uses, new jobs will be created. Horseshoe repairmen became old-school auto mechanics, who have now become diagnostic operators who predominantly replace circuit boards in cars. Almost by definition, we will be unable to conceive of the sort of labor needs that an AI-cum-robotics driven future economy may demand. It could be in interpersonal functions—such as counseling and care— that call for a unique human touch. It could be in manual, physical roles—cleaning, haircuts, massages—that machines still can’t quite replicate (at least not quickly and efficiently). If so, we may be a shift of the distribution of earnings away from university graduates, toward craftsmen and tradesmen (and women), in a rebalancing that has seen a divergence in incomes between the tertiary and nontertiary-educated workforce.

Also read: 'Don't believe AI will lead to mass job losses': Arundhati Bhattacharya, Salesforce India CEO

The corporate landscape will also be transformed. The rollout of a general-purpose technology is accompanied by firm entry and exit. This is already underway, as we see firms with greater AI exposure run ahead of those with less. Eventually, companies that fail to deploy AI extensively in their business operations will be left behind, either by competitors who are better able to do so, or because AI has rendered their entire industry obsolete. Mergers and takeovers will inevitably follow.

Ultimately, the promise of AI—or any other general-purpose technology—is that it will contribute to improved economic and societal welfare. The world has seen a dramatic slowing of productivity after the 2007/08 global financial crisis, especially in advanced economies, but also in the developing world as well. In the 1980s, Nobel prize winner Robert Solow commented that one could see the IT revolution everywhere but in the productivity statistics. The productivity boost did eventually arrive, around a decade or so later, a belated arrival that is probably due to the time it takes for technological innovations to infuse into business processes and consumer needs, and translate into meaningful changes in the way we live and work.

This is, then, the fundamental promise of the AI revolution: a revival in our anemic productivity and growth landscape. How soon we get there is unclear, as are the sorts of disruptions that will occur along the way. But it is for technology visionaries— working alongside firms and governments—to navigate the course ahead, and chart our way to a collectively brighter AI future.

Jamus Lim is an Associate Professor of Economics at ESSEC Business School Asia- Pacific.