For people who follow the development of machine learning, neural networks and artificial intelligence (A.I.), the prominence of graphics processors (GPUs) will come as no surprise; most neural networks are almost always just that — massive arrays of GPUs arranged and trained to work in concert like the neurons that make up our brains. One of the foremost manufacturers of GPUs is Nvidia, so it shouldn’t shock you that Nvidia plays an outsized role on the A.I. frontier.
Recently, though, Nvidia’s imaginativeness seems to be paying some promising (and fascinating) dividends. An excerpt from the MIT Technology Review gives us a great break down of what’s going on here:
“The software makes use of a popular new approach in AI that lets computers learn without human help. The team used generative adversarial networks, or GANs, which are neural networks that work in tandem to learn the properties of a data set.
“In a GAN, one neural network tries to produce synthetic data while another tries to tell if an example comes from the real data set or not. Feedback from the second network helps improve the performance of the first. The trick performed by the Nvidia team is to use two GANs trained on different but similar data, and to use similarities, or overlap, between the two trained models to dream up new imagery.”
For Nvidia’s use case, they had the double GAN network trained on driving condition data to imagine the same road in rain, snow, fog, etc. The ability to synthesize data is being described by some as A.I.’s learning to use their/its ‘imagination’. In Nvidia’s particular scenario, they have one of the GAN’s array imagining what the alternate driving conditions might look like. Then, the second array within the GAN is crunching massive amounts of actual data/images/video to analyze, critique, and provide feedback for its dueling GAN.
The idea is to have one GAN imagine what something might look like; the other GAN studies an ungodly amount of data giving insight into what that something actually does look like in a variety of situations. Then, by comparing the imagined image to a collection of what ought to be similar images, the second GAN can teach the first to imagine ‘better’ (read: more accurately).
Again from MIT Technology Review: “In the case of street images, for instance, one GAN was trained to internalize the properties of roads while the other was trained using images of nighttime, rainy, or snowy scenes. Connecting the two networks lets a computer imagine what a scene would look like in different conditions.”
Novel approaches like the one Nvidia demonstrated here is precisely what the nascent machine learning industry needs — innovation. Out-of-the-box thinking. True inspiration in approach. A complete change in protocol.
It’s not surprising that one of the chief enablers and empower-ers of machine learning, Nvidia, would have the vision and inventiveness to think up and subsequently build a system capable of testing this. We at ENO8 can’t wait to see what other cutting-edge firms put multiple GAN arrays up against a shared/same problem to see who and what wins, why they won and how.
Which brings us to our side chat of the day — we know A.I. is all the rage right now, especially for the media. It’s very possible the technology literally changes everything in our day-to-day lives, generally for the better. But to many, the opposite is just as plausible. It’s tough to know in a business setting if A.I. makes sense for your company, in what context and what are the potential drawbacks. Whether it’s something as benign as a voice-based U/I or as involved as IBM Watson integration, you need an innovation expert to help guide you through this new and exciting (and possibly worrying) future.
That’s where we come in.
We’re innovation experts — our job is to help you do yours better; we develop the most cutting-edge custom software and deploy it for your individual business needs. So if you’re confused or curious about how A.I. could help your business thrive, reach out via our ‘Contact Us‘ page, and we’d be happy to chat.