I guess anything that's widespread enough will make cracking it valuable enough to get someone to work around it, but it seems like we could probably come up with something that was quite a lot more human-y requiring?
Language understanding? Whatever it is that AI PHD students are failing to solve these days?
The ideal, once your email-receiving price gets high enough, would be for spammers to basically send the kind of spam that you would have paid to know about anyway (e.g. instead of reading a gadget blog, spammer pays you $1 to look at his customized spiel pitching a particular gadget similar to the ones you've previously liked).