In recent times, there has been significant attention given to the potential impact of generative AI on labor markets in the long term. It’s widely discussed how AI could render certain professions obsolete while simultaneously boosting productivity in others. However, what often goes overlooked is the current impact of AI on the labor market. This article delves into the effects of generative AI models on various professions, drawing on data from an online labor platform that reveals how AI has already affected the earnings of writers and professionals in image-related fields.
From its initial stages dating back to the 1950s, artificial intelligence (AI) has undergone a series of transformative phases. The 21st century brought remarkable advancements in AI with the rise of machine learning, neural networks, and deep learning. These innovations have integrated AI into various aspects of our daily lives, spanning communication, healthcare, transportation, finance, and entertainment. Today, AI stands at the forefront of technological progress, reshaping our routines and enhancing human capabilities. Recent developments, particularly the rapid adoption of generative AI models, have significantly improved AI’s performance, promising far-reaching impacts on the economy and society.
Traditionally, AI discussions focused on generative statistical models, which required extensive data and computational power for reliable inferences compared to descriptive models. However, the landscape has evolved with advancements in computational power and the availability of vast data sets. This evolution has propelled generative models to achieve remarkable outcomes. For instance, generative adversarial networks (GANs) have demonstrated groundbreaking abilities in generating realistic data, images, and text, particularly in pattern shifting and machine translation contexts.
One of the latest image-based generative AI models, Midjourney, has made waves with its impressive accomplishments. Midjourney can produce images based on textual and graphic inputs in various styles, showcasing its remarkable creative potential. What sets Midjourney apart is its ability to collaborate with human content creators, acting as a complementary tool rather than overshadowing the human creative process. However, it could potentially pose a threat to certain graphic engineering roles.
More recently, ChatGPT 4, a large language model by OpenAI, has garnered significant attention due to its versatility. ChatGPT facilitates natural language dialogue between humans and machines, serving not only for casual conversation but also assisting in academic research. Prominent mathematician Terrence Tao has praised GPT-based models for their assistance in mathematical research. Nevertheless, some, like AI scholar Yann LeCun, express concerns about potential misinformation. Despite these concerns, ChatGPT continues to gain users and attention, influencing the broader AI landscape.
Effects on Labor Markets: A central question that organizations grapple with is whether AI will ultimately replace or complement human workers. This question has sparked vigorous debates among policymakers and industry leaders regarding the potential benefits and challenges posed by generative AI. On one hand, AI can enhance human productivity; on the other, it may lead to job displacement and mass unemployment. The introduction of AI into organizations may also impact workers with varying skills and expertise, potentially exacerbating or mitigating wage inequality within and across occupations.
While some research has offered predictions about the future implications of generative AI on different tasks and professions, there has been relatively little exploration of its current effects. Understanding the short-term consequences and predicting long-term outcomes of generative AI in the labor market is crucial before addressing policy-related concerns.
Our Study: In our recent study (Hui et al. 2023), we provide empirical evidence on the short-term effects of generative AI on employment. Specifically, we examine the introduction of ChatGPT in November 2022 and the release of DALL-E 2 and Midjourney in April 2022. We conducted our research using data from Upwork, one of the world’s largest online labor marketplaces, known for its flexibility compared to traditional labor markets.
Our analysis, using a difference-in-differences research design, investigates how the introduction of generative AI affected freelancers in occupations heavily influenced by these AI models compared to those in less affected occupations. In the case of ChatGPT, we focus on writing-related jobs, given previous research findings and ChatGPT’s text prediction and generation capabilities. For image-based AI models like DALL-E 2 and Midjourney, we examine occupations related to design, images, and art.
Our findings reveal that ChatGPT has had a significant adverse impact on employment outcomes for workers in writing-related occupations. Freelancers in these fields have experienced a 2% decrease in the number of monthly jobs and a 5.2% reduction in monthly earnings on the platform. These effects are also observed on the extensive margin, with freelancers being 1.2% less likely to receive any job each month and, if employed, taking 4.7% fewer jobs.
For image-based generative AI models, such as DALL-E 2 and Midjourney, we have observed qualitatively similar effects, including a 2.1% reduction in the number of jobs and a 5.2% decrease in total earnings.
The Effect of Worker Quality: We also investigated whether the impact of generative AI on employment outcomes varies based on the quality of freelancers. We examined various measures of freelancers’ quality, including past employment history, earnings, skill level, performance, and hourly rates. Surprisingly, we did not find evidence that high-quality freelancers were less affected by the introduction of generative AI models. In fact, the evidence suggests that high-quality workers may be disproportionately affected by these AI models.
These findings indicate that generative AI models act as substitutes for workers across the quality spectrum, leading to reduced employment opportunities and earnings for all. These differential treatment effects align with the possibility that generative AI could narrow the productivity gap between low-quality and high-quality workers, potentially leveling the playing field.
Policy Implications: Our results carry important implications for policymakers and managers. They suggest that employers are better positioned than workers to capture the value generated by rapid AI advancements, especially as AI starts replacing human labor across various skill levels. While our study focuses on the short-term effects of generative AI, we anticipate that the substitution of technological capital for labor could become even more pronounced in the long run. With continued advancements in generative AI technology, such as GPT 4.0, the dynamics between labor and technology have the potential to be further amplified, emphasizing the need for thoughtful policies and strategies to navigate these changes.