Release Strategies and the Social Impacts of Language Models
Computation and Language
2019-11-14 v2 Artificial Intelligence
Computers and Society
Abstract
Large language models have a range of beneficial uses: they can assist in prose, poetry, and programming; analyze dataset biases; and more. However, their flexibility and generative capabilities also raise misuse concerns. This report discusses OpenAI's work related to the release of its GPT-2 language model. It discusses staged release, which allows time between model releases to conduct risk and benefit analyses as model sizes increased. It also discusses ongoing partnership-based research and provides recommendations for better coordination and responsible publication in AI.
Keywords
Cite
@article{arxiv.1908.09203,
title = {Release Strategies and the Social Impacts of Language Models},
author = {Irene Solaiman and Miles Brundage and Jack Clark and Amanda Askell and Ariel Herbert-Voss and Jeff Wu and Alec Radford and Gretchen Krueger and Jong Wook Kim and Sarah Kreps and Miles McCain and Alex Newhouse and Jason Blazakis and Kris McGuffie and Jasmine Wang},
journal= {arXiv preprint arXiv:1908.09203},
year = {2019}
}
Comments
71 pages, report