Job Cuts Due to AI: What to Make of Them?
Companies do seem to be reacting to ChatGPT and its ilk—and workers are being affected. What does this portend?
What Happened
In 2020, Sathvik Nair and I wrote that, while it was an impressive breakthrough, GPT-3 was not coming for your job. Indeed, GPT-3 did enable impressive work without causing major disruption, but ChatGPT and its ilk feel qualitatively different from their predecessors and, in the opinion of many people watching recent developments, augur badly for workers.
As the Washington Post has reported, copywriters and content writers have already started to lose their jobs because of ChatGPT. The new system, improved from GPT-3, surprised users with its helpfulness and ability to write text at a decent level. Businesses looking to cut costs saw how they could use the new tool, and some found that ChatGPT’s good-enough writing was—good enough:
Those who write marketing and social media content are in the first wave of people being replaced with tools such as chatbots, which are seemingly able to produce plausible alternatives to their work… for many companies, the cost-cutting is worth a drop in quality.
And particular workers have borne the costs of cost-cutting: Olivia Lipkin, a 25-year-old copywriter in San Francisco, found her assignments dwindling in the months following ChatGPT’s release. In April, she was let go without explanation—but from observing managers referring to her as Olivia/ChatGPT on Slack and writing that using ChatGPT was cheaper than paying a writer, she could deduce the reasons for her layoff. Eric Fein, who ran a content-writing business for 10 years, lost all 10 of his ongoing contracts that made up half of his annual income because ChatGPT was far cheaper than his wage, and clients could tolerate the quality drop.
In February, a survey found that 26% of European software and technology companies were planning to cut jobs because of ChatGPT; 22% of finance companies had plans to cut jobs for the same reason. Germany’s Bild tabloid, Europe’s best-selling newspaper, announced a major cost-cutting program and warned staff that further cuts might be necessary due to the opportunities of AI. It appears that most of these plans have not yet been carried out, though there are reports that 4000 AI-related layoffs have occurred in the US.
While its impacts aren’t massive yet, AI is an evolving technology—as its capabilities improve, the sorts of tasks that users can automate or change will increase. It’s very fair to note the limitations of current AI models, but job loss due to AI is no longer a theoretical question: we have witnessed enough of an improvement in capabilities that people are losing their livelihoods and certain jobs are changing.
Tools like ChatGPT’s new Code Interpreter, which allow the system to execute code, give us glimpses into more that might be possible—already, improvements like the code interpreter are quickly improving ChatGPT’s code-writing capabilities. Software engineers have worried about their own livelihoods, while gurus such as John Carmack have soothed worries by telling engineers that they can—and should—develop skills complementary to coding that will make them more useful and harder to automate.
The Reactions
Some predictions for the future have looked dire: Goldman Sachs economists recently predicted that the current wave of generative AI could automate as many as 300 million full-time jobs around the world. The bank estimated that approximately two-thirds of current jobs in the US and Europe are exposed to some degree of AI automation and that up to a quarter of all work could be done by AI completely.
IBM’s CEO expects AI to replace many workers and has even paused hiring to replace jobs with AI. US economists paint a more nuanced picture: many are more worried about changes in the composition of jobs than about the US running out of jobs. David Autor, an economics professor at MIT, is concerned that AI will eliminate some middle-class jobs and de-skill others, moving workers into lower-paying jobs like food service and widening inequality.
In “Why AI Will Save the World,” Marc Andreessen argued that today’s fears of job loss are echoes of a recurring panic that occurs every time a major new technology is introduced. In particular, he says “automation-kills-jobs doomers” commit the Lump of Labor Fallacy, which imagines there is a fixed amount of labor to be done in the economy at a given time. In his alternate vision, technology applied to production creates productivity growth, increased demand in the economy, new products and industries, and many other economic goods. Andreessen also makes an economic argument that the introduction of AI will reverse inequality, rather than exacerbate it, because the makers of new technologies will be motivated by market forces to drive down their prices. Andreessen’s argument is not radically different from previous rejoinders to worries about automation; for his part, he correctly observes that the “this time is different” rhetoric isn’t new, either.
Our Perspective
The phenomenon of “bullshit jobs” is not a new one, and recent AI systems have brought more attention to just how much work is truly unnecessary and automatable. But we are still at a stage where workers in many jobs will find themselves more productive when using AI systems, rather than being replaced by them. When I spoke to Shiv Rao, CEO of Abridge, we discussed how this might look for doctors. While AI scientists such as Geoff Hinton once predicted radiologists would be automated by deep learning algorithms, that prognostication remains to be proven true. Instead, doctors will likely be liberated from much of the clerical work that they have to do today because of technologies like ChatGPT—what it is to be a radiologist, and therefore what it would take to automate a radiologist’s job, will start to look totally different when these technologies start enabling them to focus on the important parts of their work.
That is to focus on one slice of a complicated story—the impact of AI systems will be heterogeneous and many people will undoubtedly be affected, even if that impact is transitory because CEOs and managers who overestimate the abilities of today’s technologies. Furthermore, AI systems are indeed improving—many are excited about the opportunities for AI agents that will more fully automate the completion of tasks.
But we’re not experiencing a sea change just yet. Job cuts are happening, and most of them still do not appear to be AI-related. Among those that are, AI may simply be an excuse for companies to cut out positions that need not exist in the first place. Numerous economic factors and headwinds are making for a difficult job market that is encouraging belt-tightening and downsizing. We should be careful not to let worries about AI-induced mass economic transformations misinform and overdetermine our understanding of what is happening now.
Conclusion
The question of automation’s impacts is not and cannot be a purely technical one. Job cuts are indeed happening, and precisely how the dynamics of the workforce evolve as AI systems grow more capable is a contingent matter. It is, of course, likely that employers will seek to extract further efficiencies from AI systems, but, as many have noted, that does not necessarily mean they will eliminate entire segments of their workforces—if humans, augmented with AI systems, can get more done, that equates to more economic value creation (of course, this will not always be the case, and plenty of employers will cut back).
Those who are employers can choose how (and whether) to integrate AI systems into their businesses, how to deal with workers whose outputs might become redundant, and other attendant concerns. Policymakers may weigh in and impose guidelines or laws to ease the burden of transition on affected workers. But we will not see a massive transition to a human-less economy overnight.