Here's a fun fact. There are approximately 365 * 24 ~ 10,000 hours in a year. If he can pull this off, then problems which currently take a year of CPU time cou

Here's a fun fact. There are approximately 365 * 24 ~ 10,000 hours in a year. If he can pull this off, then problems which currently take a year of CPU time could be solved in _one hour_ with the proviso that your results can be off by up to 1% (assuming that the error can be tracked and doesn't compound over the lifetime of the calculation). I can think of a lot of data mining/machine learning applications for which that would be a huge benefit.