HackerPapers

arXiv:2206.00169cs.LG

Discovering the Hidden Vocabulary of DALLE-2

Giannis Daras, Alexandros G. Dimakis

Early discovery. Linked by one identifiable commenter. This paper is outside the main ranking; a mention is not an endorsement.

Abstract

We discover that DALLE-2 seems to have a hidden vocabulary that can be used to generate images with absurd prompts. For example, it seems that \texttt{Apoploe vesrreaitais} means birds and \texttt{Contarra ccetnxniams luryca tanniounons} (sometimes) means bugs or pests. We find that these prompts are often consistent in isolation but also sometimes in combinations. We present our black-box method to discover words that seem random but have some correspondence to visual concepts. This creates important security and interpretability challenges.

arXiv comments
6 pages, 4 figures
Journal reference
DOI
Subjects
cs.LGMachine Learning cs.CLComputation and Language cs.CRCryptography and Security cs.CVComputer Vision and Pattern Recognition

On Hacker News

Linked by 1 identifiable commenter in 1 comment across 1 thread, from Nov 2023 to Nov 2023.

Comments linking this paper, per year.

Discussed in comments1 thread

    • It's probably just finding non-intuitive extrema in its feature space or something... the whole thing with the text in the images reminds me of this: https://arxiv.org/abs/2206.00169 and I found myself that dall-e sometimes even likes to add gibberish text unpromtedly, often…

      moritzwarhierView on HNParent context

Submitted to Hacker News

Connected through shared HN threads, not inferred similarity or a recommendation.