We are at the cusp of something truly transformational, driven by two much-hyped yet groundbreaking technology mega trends: artificial intelligence (AI) and 5G. Together they make possible things that never existed and seemed utopian not too long ago. A vivid example of this is the rise of self-driving vehicles. Notwithstanding the recent setbacks in the first iteration of self-driving, a fully autonomous self-driving vehicle will be the epitome of AI and 5G technology.
When you enter a fully autonomous vehicle, one without a driver or even a steering wheel, the vehicle does the driving for you while you sit back and relax. To make that possible, a number of complex things have to happen. The car has to continuously collect millions of data points from its thousands of sensors, dozens of cameras and multitude of other monitors. This data is fed to highly sophisticated algorithms, the “intelligence” of the self-driving AI system. These algorithms churn the data and instruct the car to drive itself safely.
If you apply the traditional AI approach to self-driving, this intelligence will reside in a centralized cloud. The data collected from the vehicle will be hauled to the cloud for processing, and instructions will be sent back to the vehicle. However, when you consider a moving vehicle in which decisions have to be made in split seconds, this approach simply won’t work. For example, if the vehicle sees any obstacle, it has to quickly determine whether it is another moving vehicle, a bike, a live person, an animal or just debris on the road – and act accordingly. The turnaround time between sensing and action must be extremely short. So, what is the solution?
Intelligent Cloud Or Intelligent Edge?
The ultimate question for any effective AI system is: Where should the intelligence lie: in a centralized cloud or in a device? In engineering lingo, devices are referred to as “edge devices” because they are at the edge of the network, while the cloud is at the center. This dichotomy is applicable to many AI applications and use cases, be it extended reality (AR/VR), medical applications, robotics, industrial, consumer, etc. The obvious choice is to keep intelligence at the edge or as close as possible to it. However, this is not practical because edge devices typically have limited processing power and are battery powered. Therefore, they cannot replicate the prowess of the cloud.
The other option would be using a fast link, like 5G, between the cloud and edge devices. However, this is not practical because edge devices generate huge amounts of data. Hauling all that raw data for trillions of AI devices would be onerous and expensive, even for 5G.
As with many things in life, the answer is a healthy middle: Move the intelligence that deals with immediacy toward the edge. Keep processing-intensive functions in the cloud. And use 5G to connect them intelligently. That’s exactly what the tech industry is working toward. It helps that tech giants such as Amazon, Facebook, Google, Huawei Intel, Microsoft, Qualcomm and others have expertise in both AI and communications.
Moving Toward Edge Intelligence
Moving toward edge intelligence simply means adopting distributed architecture for modern AI systems, wherein edge devices have the intelligence to not only collect and analyze the data that they send to the cloud, but also make crucial time-sensitive decisions. Thanks to Moore’s law, devices, be they smartphones or IoT devices, now have enough processing capability and power efficiency to run AI algorithms.
Edge intelligence in no way undermines the importance of the cloud. The cloud is and will remain a crucial part of the system. The development, training and fine-tuning of AI algorithms will still happen in the cloud. That learning is transferred to edge devices for fast decision making. This architecture can be scaled to support trillions of AI devices in the future.
Another important aspect of edge intelligence is privacy and security. It allows confidential information to be securely and privately stored in the edge instead of the cloud. This is even more important for commercial enterprises that don’t want their trade secrets to get out of their systems.
So, Why 5G?
If distributed intelligence is the perfect solution for AI and all you need is a straightforward link to transport data back and forth, why is 5G necessary? First, the link must be extremely efficient to scale up for the trillions of devices that the industry is envisioning. Second, communication between the cloud and the edge is multidimensional. Some applications, such as ones with AR/VR, will require extremely high speeds – 5G offers multi-gigabit connections. Many other applications such as drone surveillance might need huge amounts of data bandwidth – 5G brings unprecedented amounts of capacity. Others might need super-low latency – 5G offers sub-millisecond latency, which is more than 10 times quicker than 4G. Applications such as industrial controls require
extreme reliability, which 5G offers through its ultra-reliable low latency communications (URLLC) feature.
More than anything, supporting all of these applications with today’s technologies requires separate networks for each application. But with its unified air interface and networking slicing functionality, 5G can support all of them more efficiently within the same network. AND remember, with everything connected, you are looking at billions and even trillions of IoT devices that only 5G can support.
In summary, two mega technology trends of our time, 5G and AI, are joining hands to bring applications, use cases and experiences that were not possible yesterday. To make the best use of this cosmic mix, intelligence must be distributed and move toward the edge.
Technologist and wireless marketer with 20 years of experience in marketing and product management, focused on 5G, 4G, WiFi, cloud and AI.
SOURCE FORBES.COM /2018