The Benetech speaker started out making open source software for international journalists and human rights organizations to use. Then, he started working on helping blind people read books. I think that he started out with text recognition and text to speech software. 

The issue is that a computer can only read a book to someone if it sees the book as text. If it sees the book as an image, it won't know what to do with it. Thus, if a blind person buys a book, they would need to scan it in, change the images to text, and then have the software read it. Thus, he created a platform for blind people to share the books that they scanned in.

His talk made me even more frustrated with the current state of intellectual property rights. When he was working on his software for blind people, he found that there was a legal exemption to copyright law for the sake of format conversion for people with disabilities (what he was doing). His lawyer told him to ask publishers anyways. Their response: "You're the first person who has told us in advance that they were going to steal from us." He was eventually able to convince them since the law was on his side despite draconian copyright measures, since they had a legal obligation to provide electronic editions of their books for blind students (thus, Benetech would either save publishers money or prevent publishers from operating illegally), but that still provided heavy restrictions on textbooks and anything international. It wasn't until he got support from the federal government that he got any further.

His suggestion for convincing people to have less of a chokehold on their intellectual property:

"You have already made the economic decision NOT to do this social enterprise because it isn't profitable enough. We have a market that you have confirmed you are not interested in. Our idea will not hurt you -- we'll prevent it from seeping into your market. This will make you look great. Your employees will love that your product is serving a social need."

I was also surprised at how efficient they were. It costs him $50-100 to get a book ready for blind students because of automation and his peer to peer model. It costs his competitors $2000-8000. I would say that a 40-80x difference is pretty good.