Exploring the Landscape of Spatial Robustness · HackerPapers discussions

Exploring the Landscape of Spatial Robustness — mentioned by jeremyscanvic

Wed, 15 Oct 2025 08:44:31 GMT

The part about AI being very sensitive to small perturbations of their input is actually a very active research topic (and coincidentally the subject of my PhD). Most vision AIs suffer from poor spatial robustness [1], you can drastically lower their accuracy simply by translating the inputs by well-chosen (adversarial) translations of a few pixels! I don't know much about text processing AIs but I can imagine their semantic robustness is also studied. [1] https://arxiv.org/abs/1712.02779 Edit: typo

Paper and discussion history. A mention is not an endorsement.

Exploring the Landscape of Spatial Robustness — mentioned by stared

Fri, 08 Mar 2019 13:32:03 GMT

Yes, I agree with you. (No idea who downvoted my and why either.) This (well known) illusion is NOT an adversarial example. Though, I explain why for people working with AI (e.g. me) the title seemed like mentioning an adversarial example. There are plenty of examples of "just rotate and a vulture becomes an orangutan" where it does not look like an orangutan for humans. Vide: "A Rotation and a Translation Suffice: Fooling CNNs with Simple Transformations" https://arxiv.org/pdf/1712.02779.pdf

Paper and discussion history. A mention is not an endorsement.