The amazing thing about this algorithm is that it works for any data type (my other comment talks about speech and nlp). With more computers and more data it w

The amazing thing about this algorithm is that it works for any data type (my other comment talks about speech and nlp). With more computers and more data it will keep getting better - i've thought a lot about this and following this stuff for a couple of years - so it's not just some idle speculation. (I broke this recent story on hacker news a few days ago, and from here it got picked up by the new york times and spread everywhere else).