This is a brief overview of a more detailed research paper conducted as part of the AI Safety Camp. Our aim is to give those outside the economic discourse a deeper understanding of the labor market impacts of AI and to discuss potential policies in response.
AI researchers and economists broadly agree that AI has the potential to disrupt labor markets, and policymakers broadly agree that “something must be done” about it. Labor markets touch every aspect of the economy and society, and transformative AI has the potential to cause immense dislocation across countries and industries in the next few years.
However beneath this surface-level agreement lies a wide diversity of opinion. There is a notable lack of consensus in the literature for how the labor market might be affected by advanced AI in the coming decades and, in turn, how governments and regulators should respond.
We offer a straightforward framework to those outside this economic debate by which they can conceptualize the potential futures we are facing. To address this uncertainty we have written a paper that summarizes the range of labor market predictions for the future of AI in a simple model with two factors—the net task displacement effect and the aggregate productivity effect. This framework allows us to understand four potential labor markets that are possible under transformative AI. With this framework in hand we then categorize the key policy levers that are relevant under each of these different possible futures.
This framework provides a simple way to categorize and discuss predictions about the future of the labor market under transformative AI and, in turn, allows for better decision making.
Understanding AI’s Dual Impact on Labor Markets
Based on recent economic theory, we argue that AI’s influence on labor demand will depend on two primary factors: the net task displacement effect and the aggregate productivity effect. The net displacement effect measures AI’s overall impact on labor requirements by balancing task displacement and task creation in the labor market. Task displacement occurs when AI automates tasks previously done by humans, reducing labor demand. However, while transformative AI is likely to automate many tasks it will also create new tasks and jobs that don’t currently exist. The net displacement effect refers to the amount of tasks AI automates, less the tasks it newly creates.
The aggregate productivity effect captures the effect which AI will have on productivity. As AI makes jobs more efficient, production costs and consumer prices will decline. This in turn increases consumers’ purchasing power, stimulating consumer demand across the economy. To serve increased demand, firms in turn raise production, thereby curbing up the demand for labor. Thus, aggregate labor demand increases in the productivity gains AI will bring about.The relationship between these two key factors can be visualized as below, where any scenarios above the dotted line will see reduced labor demand and any scenarios below it will see increased labor demand.
These effects are not uniform and vary across different scenarios. To better understand these dynamics, we further define four potential scenarios for AI’s impact on labor markets:
Low-Effect AI: Here, AI has minimal impact on both productivity and job displacement. This scenario is plausible if AI fails to deliver transformative benefits, remaining more hype than reality.
So-So Automation: AI replaces human workers in many tasks but only marginally improves productivity. This scenario could lead to decreased labor demand, resulting in lower wages and potential job losses. As a result, labor’s share in national income is determined to fall, which would entail a rising concentration of economic wealth and political power.
Productive Synergy: AI significantly boosts productivity, while replacing relatively few human workers. Such a scenario is most plausible when AI creates a substantial amount of highly productive tasks which require human expertise. As a result, overall labor demand, wages, and employment are being boosted. Challenges mostly emanate from structural change and other adjustment costs.
Technological Supremacy: AI achieves high levels of automation and productivity, making its impact on labor demand ambiguous. The balance between displacement and productivity effects will determine whether the overall demand for labor increases or decreases. An extreme form of this scenario is the so-called “Singularity”, a point in time where machine intelligence exceeds human intelligence and economic growth accelerates.
Policy Recommendations: Steering and Adapting
To address the uncertainty ahead, we argue that policymakers should adapt a flexible and multifaceted approach. More specifically, this paper advocates for a two-pronged policy approach: steering and adaptation.
Steering Policies focus on directing technological progress to ensure strong labor demand. This includes both policies for changing what AI systems are being developed, as well as policies for influencing how the diffuse within economies:
Promoting and supporting R&D of AI-technologies which lead to labor augmentation, instead of automation.
Implementing tax incentives to encourage firms to adopt AI technologies that enhance, rather than replace, human labor.
Controlled gradual or staged release of AI-technologies.
Adaptation Policies aim to cushion the negative impacts of AI on the workforce without directly altering technological trajectories. These include:
Introducing social safety nets, such as universal basic income (UBI) or “seed UBI” programs, to provide a financial buffer for displaced workers.
Enhancing job placement services and retraining programs to help workers transition to new roles.
Implementing income redistribution measures to counteract potential increases in inequality.
Investing in education and training programs to equip workers with skills complementary to AI.
Navigating Uncertainty with Flexibility
As we stand on the brink of this technological revolution, it is crucial to remember that the trajectory of AI and its economic impact is not predetermined. Through thoughtful policy design and proactive measures, we can navigate the complexities of AI’s integration into the workforce, fostering a future where technological advancement and human prosperity go hand in hand.
This comprehensive analysis provides a framework for understanding AI’s potential impact on labor markets and offers actionable recommendations for policymakers. By embracing both steering and adaptation strategies, we can harness the benefits of AI while mitigating its risks, ensuring a balanced and inclusive approach to the future of work.
The Future of Work: How Can Policymakers Prepare for AI’s Impact on Labor Markets?
This is a brief overview of a more detailed research paper conducted as part of the AI Safety Camp. Our aim is to give those outside the economic discourse a deeper understanding of the labor market impacts of AI and to discuss potential policies in response.
The draft report can be found here. All feedback greatly appreciated.
AI researchers and economists broadly agree that AI has the potential to disrupt labor markets, and policymakers broadly agree that “something must be done” about it. Labor markets touch every aspect of the economy and society, and transformative AI has the potential to cause immense dislocation across countries and industries in the next few years.
However beneath this surface-level agreement lies a wide diversity of opinion. There is a notable lack of consensus in the literature for how the labor market might be affected by advanced AI in the coming decades and, in turn, how governments and regulators should respond.
We offer a straightforward framework to those outside this economic debate by which they can conceptualize the potential futures we are facing. To address this uncertainty we have written a paper that summarizes the range of labor market predictions for the future of AI in a simple model with two factors—the net task displacement effect and the aggregate productivity effect. This framework allows us to understand four potential labor markets that are possible under transformative AI. With this framework in hand we then categorize the key policy levers that are relevant under each of these different possible futures.
This framework provides a simple way to categorize and discuss predictions about the future of the labor market under transformative AI and, in turn, allows for better decision making.
Understanding AI’s Dual Impact on Labor Markets
Based on recent economic theory, we argue that AI’s influence on labor demand will depend on two primary factors: the net task displacement effect and the aggregate productivity effect. The net displacement effect measures AI’s overall impact on labor requirements by balancing task displacement and task creation in the labor market. Task displacement occurs when AI automates tasks previously done by humans, reducing labor demand. However, while transformative AI is likely to automate many tasks it will also create new tasks and jobs that don’t currently exist. The net displacement effect refers to the amount of tasks AI automates, less the tasks it newly creates.
The aggregate productivity effect captures the effect which AI will have on productivity. As AI makes jobs more efficient, production costs and consumer prices will decline. This in turn increases consumers’ purchasing power, stimulating consumer demand across the economy. To serve increased demand, firms in turn raise production, thereby curbing up the demand for labor. Thus, aggregate labor demand increases in the productivity gains AI will bring about.The relationship between these two key factors can be visualized as below, where any scenarios above the dotted line will see reduced labor demand and any scenarios below it will see increased labor demand.
These effects are not uniform and vary across different scenarios. To better understand these dynamics, we further define four potential scenarios for AI’s impact on labor markets:
Low-Effect AI: Here, AI has minimal impact on both productivity and job displacement. This scenario is plausible if AI fails to deliver transformative benefits, remaining more hype than reality.
So-So Automation: AI replaces human workers in many tasks but only marginally improves productivity. This scenario could lead to decreased labor demand, resulting in lower wages and potential job losses. As a result, labor’s share in national income is determined to fall, which would entail a rising concentration of economic wealth and political power.
Productive Synergy: AI significantly boosts productivity, while replacing relatively few human workers. Such a scenario is most plausible when AI creates a substantial amount of highly productive tasks which require human expertise. As a result, overall labor demand, wages, and employment are being boosted. Challenges mostly emanate from structural change and other adjustment costs.
Technological Supremacy: AI achieves high levels of automation and productivity, making its impact on labor demand ambiguous. The balance between displacement and productivity effects will determine whether the overall demand for labor increases or decreases. An extreme form of this scenario is the so-called “Singularity”, a point in time where machine intelligence exceeds human intelligence and economic growth accelerates.
Policy Recommendations: Steering and Adapting
To address the uncertainty ahead, we argue that policymakers should adapt a flexible and multifaceted approach. More specifically, this paper advocates for a two-pronged policy approach: steering and adaptation.
Steering Policies focus on directing technological progress to ensure strong labor demand. This includes both policies for changing what AI systems are being developed, as well as policies for influencing how the diffuse within economies:
Promoting and supporting R&D of AI-technologies which lead to labor augmentation, instead of automation.
Implementing tax incentives to encourage firms to adopt AI technologies that enhance, rather than replace, human labor.
Controlled gradual or staged release of AI-technologies.
Adaptation Policies aim to cushion the negative impacts of AI on the workforce without directly altering technological trajectories. These include:
Introducing social safety nets, such as universal basic income (UBI) or “seed UBI” programs, to provide a financial buffer for displaced workers.
Enhancing job placement services and retraining programs to help workers transition to new roles.
Implementing income redistribution measures to counteract potential increases in inequality.
Investing in education and training programs to equip workers with skills complementary to AI.
Navigating Uncertainty with Flexibility
As we stand on the brink of this technological revolution, it is crucial to remember that the trajectory of AI and its economic impact is not predetermined. Through thoughtful policy design and proactive measures, we can navigate the complexities of AI’s integration into the workforce, fostering a future where technological advancement and human prosperity go hand in hand.
This comprehensive analysis provides a framework for understanding AI’s potential impact on labor markets and offers actionable recommendations for policymakers. By embracing both steering and adaptation strategies, we can harness the benefits of AI while mitigating its risks, ensuring a balanced and inclusive approach to the future of work.