AI seems set to do to computer programming—and possibly other kinds of so-called knowledge work—what automation has done to other jobs, from the factory floor and the warehouse, to the checkout aisle and the call center. In those industries, the end result of widespread automation has been the elimination of countless roles—and their replacement with ones that require either relatively little skill and knowledge, or a great deal more, with workers at either end of this spectrum being rewarded accordingly.
In other words, software is eating the software industry.
Economists call this “skills-biased technological change.” It’s what happens when technology makes skilled workers more productive, while taking over the complex and difficult parts of more repetitive jobs, making workers who do them easier to train and more interchangeable.
Now, AI is automating knowledge work, and the implications for the half of the U.S. workforce who are employed in such jobs are profound. It’s true that these white-collar jobs have been evolving for decades as technology has improved, but the elimination of middle-skilled jobs seems set to accelerate as AI is institutionalized in the workplace. This new technology has the potential to reshuffle the deck of winners and losers in America’s increasingly economically polarized economy.
Coding was early to the generative AI boom that has captured the world’s attention since the release of OpenAI’s ChatGPT in November. While generative AI is typically thought of as a tool for creating text, images and even video that look as though humans created them, it’s also being used by programmers to generate code, and to automatically test it. Microsoft made GitHub Copilot—a programming tool that uses technology from OpenAI—widely available in June 2022, five months before OpenAI made public its ChatGPT bot.
Tech layoffs over the past year, driven by macroeconomic trends, happen to have come just as genuinely useful AI for coding has arrived. For many young coders, the timing is unfortunate. Data from workforce-analytics company Revelio Labs indicate that companies have tended to fire their newest employees, and that in 2023, software engineers represent the largest share of people laid off by tech companies. The few tech job openings that remain, meanwhile, are being snapped up by still-in-demand, more-experienced software engineers.
The rise of the kind of AI that can make more experienced programmers more productive could mean that companies with hiring freezes wait longer to start hiring again, and startups looking to survive the current slowdown in venture-capital investing opt not to move forward with current hiring plans.
“If I were an investor, and my companies were thinking about hiring hundreds of engineers, I’d say, well, maybe instead you can use AI to be more productive,” says Prashanth Chandrasekar, chief executive of Stack Overflow, a company that maintains a user-generated repository of questions and answers for programmers.
Stack Overflow’s latest survey of about 90,000 programmers, conducted in May, found that 70% are already using, or plan to use, AI tools in their work. Of those, a third of programmers said the number one reason they use such tools is that technology makes them more productive.
AI already helps companies do more with less
Many tech companies have promised to continue growing, despite their layoffs. Since the spring, investors have been piling into stocks perceived as being at the forefront of the AI boom, such as Nvidia, Advanced Micro Devices, Apple and Meta Platforms.
On one hand, these companies will have to continue hiring developers with skills in building AI. These engineers are currently the best-paid and the most in-demand, says Josh Brenner, CEO at Hired, a tech-jobs marketplace founded in 2013.
On the other hand, there’s Braintrust.
The worker-owned talent network represents about 360,000 freelance developers, and counts Nike, Meta and Google among its more than 1,500 clients. CEO Adam Jackson aims to double Braintrusts’ revenue in the next two years, without adding a single employee.
The company is building its own internal AI, trained on all of its job postings, developer résumés, and successful matches between the two. This system will take over more of the work of matching companies seeking developers with those freelancers, says Jackson. It will use many of the same technologies that make possible other generative AIs, like ChatGPT.
“That productivity creates more value for our network of engineers,” he says. “It came out of nowhere thanks largely to AI.”
Other industry- and company-specific AIs are likely to be built and used in a wide variety of workplaces.
Anytime you’re creating an online marketplace to match two parties—whether that’s home buyers and sellers, or ride-hailing drivers and riders—AI is a way to increase productivity without adding to your company’s head count, says Jackson.
A lost generation of early-career developers?
Many experienced developers I spoke with expressed skepticism about the ability of AI coding tools to take over the most essential tasks of programming, including designing solutions to complex problems, and understanding existing libraries of code at companies that have been building up their systems for years, or even decades.
That said, those who are already using such systems think they could eliminate the need for certain tasks that are currently generally handed off to inexperienced and early-career programmers.
“Simple off the shelf issues can be tackled relatively easily now,” says Jerome Choo, head of growth at AI-enabled, business-focused search startup Diffbot. “I can see hiring skew more senior because of this.”
It’s happened before. The job- and wage-polarizing effects of automation on an industry can interrupt the usual ladder of hiring and development, warns Yossi Sheffi, a professor of engineering at Massachusetts Institute of Technology whose latest book is on the future of work. In manufacturing, that has meant some countries—like Germany—have created a dual education system that pushes people to do both apprenticeship and university study within the typical four years of college.
With automation finally coming for knowledge work—in the form of AI—a similar rethinking of employment and education for early-career white collar workers might be necessary, he says.
“One of the main challenges of the future is how to hire junior people who don’t yet have the experience to step in when the machine doesn’t work,” says Dr. Sheffi.
In Stack Overflow’s survey of developers, there’s a difference between how experienced and junior coders said they use AI code-completion tools, like GitHub Copilot, ChatGPT, and the many startups based on it, such as Bito. While experienced programmers said their number one reason for using such tools is productivity, junior programmers said their primary use was for education—that is, to absorb what the AI already knows.
Change happens slowly
While there are early signs that AI is already disrupting the market for some developers, while enhancing the value of others, overall, technologically driven changes in demand for workers happen slowly.
When discussing the pace at which automation can push people out of a job, one of the historical examples Dr. Sheffi likes to cite is the telephone network.
In 1892, the first automatic telephone exchange was invented, says Dr. Sheffi. By 1930, America still had 235,000 telephone exchange operators. Even today, a small number of human telephone exchange operators remain.
One reason we tend to overestimate the pace of technological change in the short term is that we fail to appreciate just how many forces stand in the way of that technology becoming widespread. Those forces can include unions, government regulation and social acceptability. With today’s AI, Dr. Sheffi sees history repeating itself. Already, for example, there are proposals for a new federal agency to oversee the rollout of new technology platforms, including AI.
No matter how rapid the pace of development of AI, it’s still up to humans to adopt the technology—and in the course of individual human lives, we tend to change slowly, if at all.
That said, the double bind that many earlier-career developers currently find themselves in is a cautionary tale for us all. If AI disrupts a field at the same time that workers in it face other challenges, no matter what historians say, the impact of automation on jobs, and those who hold them, can be swift.
Write to Christopher Mims at christopher.mims@wsj.com
For more WSJ Technology analysis, reviews, advice and headlines, sign up for our weekly newsletter.
For all the latest Technology News Click Here
For the latest news and updates, follow us on Google News.