<?xml version="1.0" encoding="UTF-8"?>
<oembed>
 <version>1.0</version>
 <type>link</type>
 <provider_name>Rainbow Dash Network</provider_name>
 <provider_url>http://rainbowdash.net/</provider_url>
 <title>adiwan (adiwan)'s status on Monday, 12-Sep-22 07:38:12 UTC</title>
 <author_name>adiwan (adiwan)</author_name>
 <author_url>http://rainbowdash.net/adiwan</author_url>
 <url>http://rainbowdash.net/notice/5479784</url>
 <html>@&lt;span class=&quot;vcard&quot;&gt;&lt;a href=&quot;http://rainbowdash.net/user/42999&quot; class=&quot;url&quot; title=&quot;Anna Wright&quot;&gt;&lt;span class=&quot;fn nickname mention&quot;&gt;oracle&lt;/span&gt;&lt;/a&gt;&lt;/span&gt; I think you can make a GAN practically for everything in the realm of captchas. I think that it'll going the way I predict in order to enhance the generating models with A-B-testing (or other similar tests) by actual humans. The same as it is done now with captchas that ask humans to recognize traffic lights, cars, busses, zebra stripe crossings, bikes...</html>
</oembed>
