If you haven’t yet heard the term "deep learning," you will soon will.
It’s a branch of artificial intelligence (AI) that attempts to mimic the part of your brain where thinking occurs.
Microsoft uses it to plow through and make sense of masses of data and to power products like Skype Translator (currently in preview), personal digital assistant Cortana and Project Oxford which has tools for doing interesting things like recognize faces and expressions. (Back to the latter in a second.)
Microsoft built a deep learning toolkit called CNTK (Computational Network Toolkit) for its own purposes, but they’ve now open sourced it and made it freely available on GitHub via an MIT license.
What does this mean for enterprises? It means that they can, at least in theory, build applications that learn from their interactions with people and other computers. The best way to make sense of its capabilities is by looking what Microsoft has already done with it via its own Project Oxford.
It offers Face APIs to detect and recognize human faces in images.
So say you want to know if someone you graduated with 15 years ago is the same person who is three people in front of you in a line waiting to board a plane, take a picture with your phone and an application using the API would be able to verify (or not).
It can also guess your age with a certain level of confidence (if you’re a retailer, you don’t want to recommend wrinkle cream to a 30 year old without being tactful), your sex (we’ve all been confused by androgynous looking folks) and so on.
Not Lost in Translation
CNTK was also used to build Microsoft’s own Language Understanding Intelligence Service) LUIS, that can understand intents in language like “turn on the lights” which could be lost in translation as “turn at the street light” and so on.
In in the past something like CNTK might have been considered by Microsoft as its “secret sauce” , in this day and age, it’s something that the company is willing to share with anyone who wants to use it.
Companies who employ engineers and scientists who have the drive and know-how to work with CNTK might want to leverage it to accomplish things we haven’t experienced before, like recognize us when we walk into a store or read our mood when we call customer service.
While that might seem exploitive, it’s the very same stuff they would be able to identify a lost Alzheimer’s patient so that they can be taken home.
There are also all kinds of use cases we haven’t begun to imagine. Perhaps that’s why Microsoft has chosen to share it. Its scientists and researchers will now have a broader community to collaborate with, but not just that, it has also stepped-up to be a great corporate citizen.
Title image by modup.net