It was provided a free service[0] (search free). I'm pretty sure that no one in this conversation is saying that the cost to run the service is free, and arguing such is a strawman.
was not developed for public domain works.
pilot project was to digitize the NYT archives
While the original "recaptcha.net" website is no longer available online (even via archive.org). There are plenty of sources still available that bely this claim. that will help convert printed text into computer-readable letters on behalf of the Internet Archive[0]. The team is involved in digitising old books and manuscripts supplied by a non-profit organisation called the Internet Archive[1]. "There's still about 100 million books to be digitised, which at the current rate will take us about 400 years to complete - Luis von Ahn, Carnegie Mellon"[1]. The sources are from 2007, when recaptcha was first introduced. There are more, but I picked the first two by going back to the beginning of the wikipedia entry for recaptcha and looking at the supplied sources.
This is before Google bought it in 2009. While I can't speak for anyone else, this is what I mean when I say "original" recaptcha.
laugh at your suggestion that Google could "pay a group of people to do".
They got a group of people to do it for free, which implies (barring salaries) that they could also get a group of people to do it for not-free. Thus the argument that people don't want to work for Google for free.
Google hadn't learned from massive user data
And when Google has met it's ML goals and decides it gets no further benefit from recaptcha? Remember, it's not a free-to-run service. They only continue it as long as they benefit. If recaptcha shuts down, it would have been nicer to have the fruits of that work available through something like the Internet Archive, than something like Google Books.
[0] - http://www.cmu.edu/news/archive/2007/May/may24_recaptcha.sht...
According to Wikipedia and the New York Times, reCaptcha was not developed for public domain works. Its pilot project was to digitize the NYT archives, archives which were not released to the public domain nor are fully available without being a subscriber: http://www.nytimes.com/2011/03/29/science/29recaptcha.html
I'm not a machine learning expert but I'm going to laugh at your suggestion that Google could "pay a group of people to do". In the above referenced NYT article from 2011, recaptcha's creator says several million words were being processed by recaptcha per day.
And again, have to disagree that the end user "does not benefit in any significant way". We would not be discussing this if Google hadn't learned from massive user data to iterate their captcha from distorted word mush to what it is today. Captcha was a serious drain of user energy and patience, that's why recaptcha was invented in the first place. And the worst captchas were tolerated because automated usage was a financial threat to websites that end users use.