Originally Posted by
krazykanuck
I feel like "machine learning" is just the buzzword of the last year or two. I hear it so much at work, I'm like guys, this is no different than what we did before this term was invented. But whatever.
Except that before we didn't have computing power to perform more sophisticated tree searches and ability to handle large data sets in reasonable time. The underlying statistical methods are more or less the same, ability to run complex simulations is not. The range of objectives set for the "two engineers" (Beavis and Butthead?) is too large to meaningfully speculate about, since you don't need machine learning to just weed out "gamers", but you might need it to figure out their net effect in a larger context. In other words, AMEX might be losing on gamers in one domain but gains more in another, which is a more complex problem to figure out.