Product design and development firm Cambridge Consultants developed a deep learning-based system that turns human sketches into paintings that resemble Van Gogh, Cézanne and Picasso. "What we've built would have been unthinkable to the original deep learning pioneers," said Monty Barlow, director machine learning at Cambridge Consultants in reference to their interactive system that call Vincent. "By successfully combining different machine learning approaches, such as adversarial training, perceptual loss, and end-to-end training of stacked networks, we've created something hugely interactive, taking the germ of a sketched idea and allowing the history of human art to run with it." Once trained on nearly 200 million parameters, Vincent is able to understand the important edges in paintings and uses this understanding to produce a complete picture.
The issue lies with a prevalent tactic in AI development called "back propagation". Geoffrey Hinton has been called the "Godfather of Deep Learning". It relates directly to how AIs learn and store information. Since its conception, back propagation algorithms have become the "workhorses" of the majority of AI projects.
NVIDIA GPUs have been on the forefront of accelerated neural network processing and are the de facto standard for accelerated neural network research and development (R&D) plus deep learning training. At the NVIDIA GPU Technology Conference (GTC) in Beijing China earlier this week, the company maneuvered to also become the de facto standard for accelerated neural network inference deployment. At GTC Beijing, NVIDA lined up the major Chinese cloud companies for AI computing: Alibaba Cloud, Baidu Cloud, and Tencent Cloud. At GTC-Beijing, it announced inference designs with Alibaba Cloud, Tencent, Baidu Cloud, JD.com, and iFlytek.
NVIDIA's meteoric growth in the datacenter, where its business is now generating some $1.6B annually, has been largely driven by the demand to train deep neural networks for Machine Learning (ML) and Artificial Intelligence (AI)--an area where the computational requirements are simply mindboggling. First, and perhaps most importantly, Huang announced new TensorRT3 software that optimizes trained neural networks for inference processing on NVIDIA GPUs. In addition to announcing the Chinese deployment wins, Huang provided some pretty compelling benchmarks to demonstrate the company's prowess in accelerating Machine Learning inference operations, in the datacenter and at the edge. In addition to the TensorRT3 deployments, Huang announced that the largest Chinese Cloud Service Providers, Alibaba, Baidu, and Tencent, are all offering the company's newest Tesla V100 GPUs to their customers for scientific and deep learning applications.
For example, a company called Intelligent Flying Machines built a drone that can autonomously navigate through a warehouse and match what's on the shelves to what's in the inventory system to help the distribution center manage inventory better. We see a lot of opportunity in other areas like precision agriculture, package delivery, safety and security, and search and rescue. For the areas that I mentioned -- industrial inspection, precision agriculture, package delivery, safety and security, search and rescue -- there's going to be an opportunity for UAVs to solve these challenges in a way they haven't been able to before. Clayton: Nvidia makes Jetson, and Jetson is Nvidia's platform for artificial intelligence for edge devices like UAVs.
The story behind the story: a finely tuned generative adversarial network that sampled 8,000 great works of art -- a tiny sample size in the data-intensive world of deep learning -- and in just 14 hours of training on an NVIDIA DGX system created an application that takes human input and turns it into something stunning. Building on thousands of hours of research undertaken by Cambridge Consultants' AI research lab, the Digital Greenhouse, a team of five built the Vincent demo in just two months. After Huang's keynote, GTC attendees had the opportunity to pick up the stylus for themselves, selecting from one of seven different styles to sketch everything from portraits to landscapes to, of course, cats. While traditional deep learning algorithms have achieved stunning results by ingesting vast quantities of data, GANs create applications out of much smaller sample sizes by training one neural network to try to imitate the data they're fed, and another to try to spot fakes.
A test fleet of autonomous delivery trucks scheduled for deployment next year will be outfitted with self-driving system based on Nvidia's AI processing technology for autonomous vehicles. The 2018 demonstration will use the package delivery company's (ETR: DPW) fleet of 3,400 electric delivery vehicles outfitted with cameras, radar and lidar (light detection and ranging). "The development of autonomous delivery vehicles demonstrates how AI and deep learning are also reshaping the commercial transportation industry," Nvidia CEO Jensen Huang noted in a statement announcing the partnership during a company event this week in Munich, Germany. Drive PX combines deep learning, sensor fusion and machine vision with the ability to figure out location using onboard maps to plot a safe course.
This was followed by the implementation of NVIDIA DGX-1 systems with NVIDIA Tesla P100 graphics processing units (GPUs) in SAP's production data center in St. Leon-Rot, Germany and in SAP's Innovation Labs in Palo Alto, California, and Singapore in September 2017. From the outset of SAP's machine learning efforts, NVIDIA's computing platform has promoted the company's training of data sets and algorithms – the core of intelligent machine learning applications in the SAP Leonardo Machine Learning portfolio. With SAP Leonardo Machine Learning, SAP brings digital intelligence to enterprise offerings and creates tremendous opportunities for customers to realize greater benefits through automated processes, targeted results-driven marketing, superior customer service, as well as increased agility and process efficiency. The partnership between SAP and NVIDIA to bring DGX-1 systems with Volta to production in the SAP Data Center will give SAP customers access to machine learning services and applications from SAP's own Data Center infrastructure.
Just in time for the fall sports season, researchers are developing an AI-powered app that detects concussions right on the playing field. Working with a team of UW researchers and clinicians, he is using GPU-accelerated deep learning to create an app that detects concussions and other traumatic brain injuries with nothing more than a smartphone camera and 3D-printed box. The app, called PupilScreen, assesses the pupil's response to light almost as well as a pupilometer, an expensive machine found only in clinical settings. In a pilot study of 42 patients with and without traumatic brain injury, the app tracked pupil size almost as well the pupilometer.
By the middle of 2018, Nvidia believes it will have a system capable of level 5 autonomy in the hands of the auto industry, which will allow for fully self-driving vehicles. Pegasus is rated as being capable of 320 trillion operations per second, which the company claims is a thirteen-fold increase over previous generations. In May, Nvidia took the wraps off its Tesla V100 accelerator aimed at deep learning. The company said the V100 has 1.5 times the general-purpose FLOPS compared to Pascal, a 12 times improvement for deep learning training, and six times the performance for deep learning inference.