On one hand, I understand (and largely agree with) the various concerns about algorithm-izing students, grade inflation, and using the money directly to fund teachers. Our public education system has a shit-ton of critical P1 bugs.
However. On the other hand, I'm reminded to Neal Stephenson's "Diamond Age" interactive primer. First, the manifestation from the book is clearly not the initial beta version; some lengthy amount of revision and improvement would clearly be necessary to achieve the level of sophistication presented in the storyline. As such, Gates' initial investment along these lines could be a key first step down this path. Second, the manifestation from the book demonstrates a clear power over educational thought that is conveyed by the approach -- that is, if creating an algorithm for more effective education WAS attainable, then it would be an algorithm which could single-handedly influence entire generations* down a path of compliance with, or rejection of, status quo systems.
* This technology would also likely be available only to the elites with money to fund it, at least initially. Perhaps GNU GPL versions would surface over time. Given the power over educational philosophy, would the developers have to be thoroughly licensed in the educational system first? With versions of the software that are considered with the same distaste as allowing middle-schoolers to read Hunter S Thompson?