Work

Speak, Memory: An Archaeology of Books Known to ChatGPT/GPT-4

Year: 2023

Type: conference-paper

Abstract: In this work, we carry out a data archaeology to infer books that are known to ChatGPT and GPT-4 using a name cloze membership inference query.We find that OpenAI models have memorized a wide collection of copyrighted materials, and that the degree of memorization is tied to the frequency with which passages of those books appear on the web.The ability of these models to memorize an unknown set of books complicates assessments of measurement validity for cultural analytics by contaminating test data; we show that models perform much better on memorized books than on nonmemorized books for downstream tasks.We argue that this supports a case for open models whose training data is known.

Source: Conference on Empirical Methods in Natural Language Processing (EMNLP)

DOI: 10.18653/v1/2023.emnlp-main.453

Authors: Kent Chang, Mackenzie Cramer, Sandeep Soni, David Bamman

Institutions: Emory University, University of California, Berkeley

Language: English

Topic: Topic Modeling

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: hybrid

Funders: National Science Foundation, National Endowment for the Humanities

Awards: CAREER: Using Fiction to Improve Real-World Information Systems, Multilingual BookNLP: Building a Literary NLP Pipeline Across Languages