Modern AI for image recognition is easily fooled by signatures

Researchers at OpenAI have found that their state-of-the-art computer vision system can be easily fooled

If you label an object on top of it, the artificial intelligence (AI) model recognizesits wrong.

Researchers at OpenAI call this typographic attacks. They use the model's ability to reliably recognize text. It turns out that even a photo with handwritten text can fool AI.