OpenAI Economic Impacts Research Initiative

OpenAI is establishing a collaborative research framework to measure the economic impacts of large language models (LLMs) on individuals and society. This initiative aims to generate high-quality evidence to inform AI deployment policies, system design, and public policy interventions to maximize human prosperity and minimize economic harm.

Objectives of Economic Impact Research

OpenAI seeks to develop rigorous tools for measuring how AI models affect the economy to ensure that artificial general intelligence (AGI) benefits all of humanity. The organization identifies three primary axes where this research will influence decision-making:

  • AI Deployment Policies: Determining how and when models are released to the public.
  • AI System Design: Making technical decisions during the development of AI systems based on their predicted economic outcomes.
  • Public Policy Interventions: Providing real-world evidence to help governments and institutions create policies that mitigate risks.

By studying current models such as Codex, DALL-E 2, and ChatGPT, OpenAI intends to create a foundation for assessing the safety and economic viability of more capable systems in the future.

Context on Economic Risks and Inequality

OpenAI notes that while there is currently little immediate risk of widespread technological unemployment due to AI, the labor market impacts will vary significantly across different types of workers. The company highlights several systemic concerns:

  • Uneven Distribution of Prosperity: Historical trends in the US over the last forty years show a worrying pace of economic divergence and uneven access to opportunity.
  • Wellbeing Impacts: Even transitory unemployment shocks can have widespread negative effects on individual wellbeing.
  • Societal Cleavages: Increasing economic inequality has the potential to amplify existing societal divisions.

Collaboration Framework for Researchers and Companies

OpenAI is prioritizing external partnerships over exclusively in-house research to broaden the evidence base. The initiative provides two distinct pathways for participation:

For PhD-Level Researchers

OpenAI is seeking expressions of interest from PhD-level researchers and current doctoral students. Potential collaborations may include:

  • Integration into a larger research cohort.
  • Participation in red teaming exercises.
  • Provision of API credit grants.
  • Future connections to customers building LLM-based products to study direct impacts on workers and consumers.

Researchers are evaluated based on their background, the clarity of their motivation to collaborate, and the decision-relevance of their research interests.

For Companies and Users

OpenAI is creating a portal for firms and users of OpenAI tools to contribute to or sponsor research. The goal is to facilitate discussions between external researchers, AI developers, AI-adopting firms, and affected workers to incorporate a wider range of perspectives into the path of AI development.

Sources