If the article is correct and AI has a hard time generating or understanding optical illusions, could they be exploited to determine human from computer?<p>After typing that out, I'm not sure I could actually answer a question about the referenced optical illusion [1] in the article. For example, what direction is the full wheel in the upper-left turning? Who knows, but I have a headache now.<p>Edit: it seems this has been discussed before [2], [3], [4]<p>[1] <a href="https://i.kinja-img.com/gawker-media/image/upload/u2fqwmggufhl6edenlna.jpg" rel="nofollow">https://i.kinja-img.com/gawker-media/image/upload/u2fqwmgguf...</a><p>[2] <a href="https://security.stackexchange.com/questions/116612/using-optical-illusions-as-captcha" rel="nofollow">https://security.stackexchange.com/questions/116612/using-op...</a><p>[3] <a href="https://arstechnica.com/civis/viewtopic.php?f=23&t=59240" rel="nofollow">https://arstechnica.com/civis/viewtopic.php?f=23&t=59240</a><p>[4] <a href="https://qz.com/1047988/amazon-amzn-has-a-new-captcha-that-youre-designed-to-fail/" rel="nofollow">https://qz.com/1047988/amazon-amzn-has-a-new-captcha-that-yo...</a>