Over the past few weeks, Nvidia has been holding a series of regional GPU Technology Conferences (GTC) in different parts of the globe. In September, Nvidia showed off its new Tensor3 GPU made for artificial intelligence (AI) inferencing in China. This week, the company took its show to Munich to host GTC Europe where it made a couple of announcements in the advancements of self-driving vehicles.
The quest for the fully autonomous car has been somewhat of a “holy grail” and one of the best examples of what’s possible when discussing advanced technologies such as machine learning, artificial intelligence (AI), and the Internet of Things (IoT).
Test fleet of autonomous delivery vehicles
Nvidia CEO Jensen Juang made two self-driving announcements during his keynote. The first one was made with Deutsche Post DHL Group (DPDHL), the world’s biggest mail and logistics company, and ZF, an automotive provider. The three companies have teamed up to deploy a test fleet of autonomous delivery vehicles (pictured above), starting next calendar year.
DPDHL will deploy electric light trucks powered by the ZF ProAI self-driving system, which is powered by the Nvidia DRIVE PX palm-size supercomputer but also includes sensors, cameras, LIDAR and radar that feed the data into the system. The AI-based system will automate the process of transporting and delivering packages from the central point to the final destination, which is the most expensive aspect of courier services. The autonomous vehicles will be able to use the AI to comprehend the environment it is in, then plan a safe path and park itself. It ensures safe, accurate deliveries at a significantly lower cost than using people.
It’s important to understand that self-driving vehicles require compute resources in the vehicle, but also data center resources to do the training for the AI to learn all the objects it will encounter while moving about a city. Earlier this year, I pointed out that GPU computing was one of the future data center technologies IT leaders should keep an eye on. For DPDHL, the future is now as it deployed the GPU-powered, Nvidia DGX-1 AI supercomputer to do the training. As other verticals start to rely more heavily on machine learning and AI, GPUs will become a bigger part of their data center strategies.
New level of autonomous vehicles
The second announcement takes robo-taxis to the next level, level 5 to be exact. There are many different levels to self-driving cars. Level 0 is no automation. Level 1 has basic features, such as cruise control, then it escalates from there. Level 5 is full automation where the car drives full time under all conditions.
At the event, Nvidia introduced its DRIVE PX AI platform (code-named Pegasus), which can handle over 320 trillion operations per second, which is 10x the performance of its previous system, the DRIVE PX 2.
The new Drive PX will introduce a new level of autonomous vehicles. Existing ones, such as the popular Tesla, still retain the look and feel of a car with a steering wheel, driver seat, etc. With level 5, think of the vehicle as being more of a moving lounge. There will be no steering wheel, pedals, mirrors or anything else a driver may need.
When solutions like this become reality, and they will, business productivity could go through the roof. Instead of wasting time in a car commuting, people can use that time to work or even hold meetings. Level 5 vehicles aren’t only for the busy worker. Elderly people or disabled people could gain an entirely level of freedom because the car could pick them up and take them to appointments instead of having to put the burden on a relative.
Fully autonomous vehicles can also save lives because people will never fall asleep at the wheel or get in accidents from texting and driving. I know many people have the attitude of never trusting an autonomous vehicle, but with the right data and AI, self-driving cars will make travelling by car much safer. Personally, I can’t wait because I can use the hour it takes me to get to the airport to write even more Network World blog posts.
Similar to the delivery vehicles, much of the ongoing learning will be done in big data centers, which will calculate massive amounts of data. The rules learned there will be executed in the car by an onboard system, such a Pegasus.
Autonomous vehicles are very much the shape of things to come regarding other digital initiatives. Not all data will be processed at the edge nor will it be done in public or private clouds. Digital technologies require everything to be connected, hence the big IoT push by many companies, but also lots and lots of data needs to be processed where it makes sense. In the case of an autonomous vehicle, much of the initial learning rules can be done in a data center, but the decision of whether to stop or not needs to be done in the car.
IT leaders should consider this “cloud-to-edge” strategy as they plan out future initiatives. Technologies such as IoT, AI and mobility are changing the world faster than ever before. The key is to have the right compute systems in the right places to make the best decisions in as short a time as possible.