Making Intelligence A Bit Less Artificial
from the making-AI-useful dept
There have been plenty of stories on the topic of recommendation engines gone awry (such as the ever popular story of those who own TiVos that think they’re gay). New research is suggesting that most people don’t get very good recommendations from those types of artificial intelligence systems. Those systems don’t seem to understand context all that well. I know I’ve had problems where Amazon recommends odd books for me because I once bought a friend a book on some random subject. So, the latest move is to take the “artificial” part out of artificial intelligence and hand the work over to human intelligence, instead. Companies are looking at employing “category experts” or other humans who act as filters on top of the recommendation engines. I’ve always been a fan of business models that involve both a technology component and a human component – since I think they’re both more realistic and more defensible from a competitive standpoint.