Ever since reCAPTCHA become mainstream we knew that Google was using us as mechanical turks. At the beginning it seemed OK because we were reading books and newspapers, but then the system started asking questions like which of these squares had cars or traffic signs or store fronts in it, and you had to be kinda dumb not to realize you were training an Street View AI. Since then, I’ve been trying to avoid sites that ask me to work for free for Google.
On the other hand, I’m kinda sick and tired of the SPAM that comes through the RSS feeds of https://blogs.osm.org/ . I think this is mostly due to the fact that opening an account does not involve proving you’re not a robot.
So, what about creating a captcha system that at the same time helps building the OSM database? What is not clear to me is what exactly the system would be helping to do. I think the simplest and more impacting would be to recognize simple features that would help mappers to find things to maps, more than trying trying to train an AI to map automatically.
I wonder what would the suppliers, like Bing, Mapbox, etc, think of such use of the imagery they license to OSM for mapping, specially if the system starts to be used by third party sites to block bots/spammers.
 If you want to know a little more, please read https://www.techradar.com/news/captcha-if-you-can-how-youve-been-training-ai-for-years-without-realising-it.