i think the time for you keystroke lottery idea is over. Computers are more and more powerful and making intelligent decisions now. probably can already read lots of handwriting and will get better. Computerized translations also getting better and better. So why use all those people by turning typing into a lottery?  It can only be transitional. Also, mechanical turk (amazon)  already does a lot of typing. That will only get bigger. I'm not saying you lotteries can't work, just they will probably becoming uneconomical in the face of computers faster and cheaper.

Computers will definitely get more and more powerful. But how smart they will get depends on the problems they are given to solve. Consider self-driving cars. The more of them on the road the safer they and everyone else will be because they will communicate and coordinate with each other. In so doing they will achieve logical consensus on what to do in any given situation. Currently self-driving cars are so few in number they can only react, not coordinate. To coordinate requires a whole new and massive body of data to work with. So conventional cars must begin coordinating with each other too until the driver becomes an actual hindrance.

Developers of handwriting recognition and translation software will probably get around to developing KLs themselves as an economical data-gathering technique to improve their own product. They could scan a bunch of handwritten pages from a library archive (the way Google currently scans books); submit those pages to the world via a KL; then compare those results with the software they have already developed; and then tweak or butcher their algorithms to produce the KL result. KLs could turn out to be the essential tool in the development of genuinely effective handwriting recognition and translation: KLs as AI’s killer app. 

If anything is transitional, it would be Amazon’s Mechanical Turk (and maybe Gigwalk.com). MT doesn’t invite workers by the millions as a KL would. It ruthlessly rejects them by the millions with its punishingly low pay for work done. Those who accept that pay must be hyper-alert to new postings, start on time and finish the task within deadline. In the case of typing, you will need a proofreader who will work for the same pay and deadline. You don’t get consensus that way. KLs offer massive redundancy to achieve the best consensus possible for a given document, page, paragraph, word, letter, symbol, or digit and (possibly) translation. With the lottery model, there is no deadline or schedule. Anyone could just log on and type as little or as much as they pleased, trying to achieve consensus with as many others as possible, because only a consensus-validated keystroke can win the payout. Conceivably a lot of work being done by MT would migrate to KLs. When you think about it, MT and Gigwalk (which probably could be turned into a lottery for its strolling data-collectors) are unimaginative, conventional attempts to crowdsource certain kinds of work. They would probably be a lot more successful if turned into a way for their participants to trade their time and bits of work for a Lotto-sized payoff, at least for certain kinds of tasks.

As far as I can see, without KLs the only way computerized handwriting recognition and translations could ever outperform people would be if handwriting were written out in the first place by another computer using a known algorithm; and if people spoke languages designed by computers too. Until then, or until KLs become a reality, the old printshop maxim will apple: You can have your job cheap, fast, and right — but only two out of three. Faster computers alone won’t change that; MT will not change that. But KLs probably could change that.

Worth a try anyway. Read my full posting on all this at: https://www.scribd.com/doc/254106778/Keystroke-Informant-Lotteries-A-Bitcoin-DAC-Killer-App). It’s a lot more readable than the Slideshare version, which doesn’t include my description of a hypothetical Informant Lottery (currently on page 11).