Understanding the Capabilities, Limitations, and Societal Impact of Large Language Models

TL;DR

This paper summarizes a multidisciplinary discussion on GPT-3, exploring its technical capabilities, limitations, and societal impacts, highlighting the need for responsible development and deployment of large language models.

Contribution

It provides a comprehensive overview of expert insights into the technical and societal aspects of large language models like GPT-3, based on a diverse research community's discussion.

Findings

Large language models have significant technical capabilities.

There are notable limitations and risks associated with these models.

Societal impacts include ethical, political, and communication challenges.

Abstract

On October 14th, 2020, researchers from OpenAI, the Stanford Institute for Human-Centered Artificial Intelligence, and other universities convened to discuss open research questions surrounding GPT-3, the largest publicly-disclosed dense language model at the time. The meeting took place under Chatham House Rules. Discussants came from a variety of research backgrounds including computer science, linguistics, philosophy, political science, communications, cyber policy, and more. Broadly, the discussion centered around two main questions: 1) What are the technical capabilities and limitations of large language models? 2) What are the societal effects of widespread use of large language models? Here, we provide a detailed summary of the discussion organized by the two themes above.