Netflix's movie recommendations are already eerily accurate, but they're still no match for an actual person who knows what you do and don't like. However, a developing style of computing may be able to change that by simulating a human brain across thousands of machines — and now Netflix has begun trying it out. In a blog post, Netflix describes how it's begun experimenting with neural networks and deep learning to see if they might be able to improve its personalization features.
It's still far, far out before Netflix could make use of these techniques, but it appears to have had success in getting the research started. Rather than investing in its own expensive data center to run the trials on, Netflix instead utilized Amazon Web Services, which allowed it to rent the infrastructure it needed for its fairly small-scale experiment. For now though, deep learning algorithms still have plenty to learn: while they might be able to pick up patterns on their own, they're still quite basic and can't pick objects out all that reliably.