TMCnet - World's Largest Communications and Technology Community

TMCnet Feature

February 12, 2014

Netflix Working on Artificial Neural Networks to Make Better Recommendations

One of the few good things that could be said about the video store was that, under the right circumstances, there was a video store clerk on hand who knew a good deal about movies. This could be especially helpful for those who were looking for a certain type of movie but weren’t sure where to start. But as the video store's time waned, and the rise of online sources like Netflix stepped in to take its place, that video store clerk was largely lost to algorithms and automated systems. We're still a long way from video store clerk-grade quality, but that's something Netflix wants to change.

To that end, a recent blog post emerged describing work on what was described as an artificial neural network, comprised of large quantities of graphics processing unit (GPU) chips, as opposed to using full-on central processing unit (CPU) chips that would have proven much more expensive and time-intensive to build. With the use of GPUs—most of which at last report were Nvidia-make—running on the public-cloud infrastructure offered up by Amazon Web Services, the resulting research could at least approximate a video store clerk, so to speak.

Image via Shutterstock

Though Netflix is keeping comparatively quiet on the overall purpose of the research so far, there are certain conclusions that crop up almost immediately. First, this is a boost to Nvidia, really; Netflix has taken an array of its GPUs and is putting said devices to work on a project of almost staggering scope. Some might well turn to Nvidia GPUs in light of such a move, which really only bolsters Nvidia's fortunes. It also puts Netflix's cloud strategy clearly on display, and gives Amazon a boost as well in that sense. But what really jumps out on this one is the connection to big data.

One of the biggest things about big data is that it's not enough to just have a whole lot of data on hand. Yes, it's absolutely necessary to have that data, but just having the data isn't enough. Netflix has ridiculous amounts of data; just check out the front page of an account some time to see what's big on Facebook, or what's big in a certain region, or what's big in a certain genre. That's just the start of things, too. The only problem here, of course, is that all that data has to make sense somehow. The data has to be searched for patterns. Patterns have to be compiled to see if those patterns turn into trends. If there are trends, the trends have to be acted upon. There are points all up and down the dial to make something happen here, and Netflix may be in the process of making a network capable of understanding these points and providing a way to act on same.

It's an exciting principle, though only time will tell just how far it all goes. It could be that Netflix is about to bring in some powerful new features that will make suggestions a lot closer to the mark in terms of what movie someone might like to watch based on that person's past viewing history. Still, this could be a very exciting time ahead for those who stick to Netflix for online viewing, with the equivalent of an always-on video store clerk ready to say just what's going to be most likely worth watching.

Edited by Alisen Downey

View all articles

Comments powered by Disqus

Related Articles



Technology Marketing Corporation

800 Connecticut Ave, 1st Floor East, Norwalk, CT 06854 USA
Ph: 800-243-6002, 203-852-6800
Fx: 203-866-3326

General comments:
Comments about this site:


Subscribe to our FREE eNewsletter


© 2014 Technology Marketing Corporation. All rights reserved.