8
[2305.00118] Speak, Memory: An Archaeology of Books Known to ChatGPT/GPT-4
source link: https://arxiv.org/abs/2305.00118
Go to the source link to view the article. You can view the picture content, updated content and better typesetting reading experience. If the link is broken, please click the button below to view the snapshot at that time.
Computer Science > Computation and Language
[Submitted on 28 Apr 2023 (v1), last revised 20 Oct 2023 (this version, v2)]
Speak, Memory: An Archaeology of Books Known to ChatGPT/GPT-4
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 non-memorized books for downstream tasks. We argue that this supports a case for open models whose training data is known.
| Comments: | EMNLP 2023 camera-ready (16 pages, 4 figures) |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2305.00118 [cs.CL] |
| (or arXiv:2305.00118v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2305.00118 |
Submission history
From: Kent Chang [view email][v1] Fri, 28 Apr 2023 22:35:03 UTC (6,906 KB)
[v2] Fri, 20 Oct 2023 21:23:21 UTC (44 KB)
Recommend
About Joyk
Aggregate valuable and interesting links.
Joyk means Joy of geeK