Artificial intelligence, quantum computing, neural networks, Big Data across connected systems. All are different and at the same time related to a target, the get machines to acquire the ability to think as human beings do. And the Deep Learning is one of the steps necessary to achieve this.
We do not mean that we are able to create Skynet so then it rebels against humanity but to develop systems with advanced intelligence, systems with an equivalent to the human intelligence to be able to process vast amounts of data from a future interconnected world. Qualcomm has something to say about it.
The clouds in your Pocket
One of the keys in the development of the artificial intelligences passes by unsupervised learning systems to provide. In Christian, that systems are able to learn by themselves, without any interaction or human action during the process. In this learning appeared some time ago a fundamental concept: the Deep Learning.
This capability of approaching learning systems or machines human perception has been supported so far in the cloud. It makes sense, because the best way to get data from multiple sources is to obtain them from a common link, a server connected to the internet. Qualcomm intends to now get rid of that necessary connection and move the Deep Learning into the heart of your smartphone.
It has trick, clear is, as Qualcomm will force necessarily go through one of their processors. Specifically, the series of processors from high-end initiated with the Snapdragon 820, presented at the end of 2015 and running already in 2016, and continue with the 823, 828 and 830. The latter already for next year.
Zeroth, a neural processing engine
Qualcomm wants to get it through a new SDK developed by the Americans and known as SDK Zeroth, or “Qualcomm Snapdragon neural processing engine”. A pretentious name for a project as complex as promising, move the deep learning, unsupervised learning of systems to your smartphone.
The reasons for Qualcomm to tackle this project can be summed up in two simple concepts: Privacy and speed. Offer this level of processing and learning to users of its SDK allows high levels of privacy, run by and for the smartphone, without intermediation from the outside, and also a very low latency to not require a transfer of information.
Zeroth means better privacy and one lower latency not having loads of data to the cloud. (In relation to a medical application that studies skin with the smartphone’s camera) Much more sense to carry out the classification of images on the device itself had, and no matter what kind of data are processed. They may be Visual data, it could be audio.
A SDK to build apps based on Deep Learning
Will be limited, of course, but Qualcomm will provide companies that embrace their SDK the capacity of build own and mobile applications of Deep Learning, Analytics that will not require disbursement of large and powerful servers to run the entire process inside a tiny processor as that carry mobile phones.
Qualcomm already has support in this project, companies like Nauto which are already using the SDK in tests to produce cameras integrated into vehicles softwares. Thanks to these learning skills studying the behavior of the driver, the rest of the traffic and cameras that can provide a large amount of data when it comes to detailing an accident. Or prevent it.
With the advent of technologies such as Deep Learning and cognitive computing the way we learn, relate and understand the world will also change radically. And Qualcomm wants his piece of cake in this evolution of technology, you want to be one of those who brought her to the smartphone.