causal

Neural Networks Are Easily Fooled By Context Shifts

Machine learning systems are easy to fool because big data causes them to focus on statistical features — like textures rather than edges, or a fire truck's typical orientation — that humans don't rely on, so small changes such as photoshopping a fire truck upside down cause complete misclassification despite humans recognizing it instantly.

causalpending

Speaker

Melanie Mitchell

Evidence Quote

“when we say we've trained them to recognize fire trucks, it's not totally clear what we've actually trained them to recognize.”

Source

68 | Melanie Mitchell on Artificial Intelligence and the Challenge of Common Sense— Sean Carroll's Mindscape
Created: 6/13/2026, 12:06:58 AM

Your notes

Loading notes...