There is a lot of buzz in the technology sector around AI and 5G and their transformative potential. What is interesting about both technologies is that they are universally applicable across many industries and are likely to change the way many of them operate. For that matter, each technology has its own set of use cases, which are also inextricably connected to one another. The pairing of 5G and AI stands to amplify the rate and impact of both technologies. Covid-19 has accelerated the need for both technologies even further, due to their potential to improve video conferencing, network capacity and more. While my colleague Will Townsend recently looked at the juncture of 5G and AI in infrastructure, today I’m mostly going to look at it in terms of the edge and client devices.
5G and AI are better together
Due to the broad industry applications of 5G and AI, there are many places where the two technologies overlap and benefit from one another. One of the reasons why AI and 5G are so complementary is that AI exists on a continuum of compute—it can be on-device, at the MEC (multi-access edge compute), edge cloud, cloud or data center. 5G provides the powerful, reliable connectivity necessary to maximize the intelligence and responsiveness of AI across all these network layers and client devices.
Conversely, AI can help 5G minimize the amount of data that is generated by client devices, so that they only send pertinent data back to the cloud. There are many 5G and AI players in the IT space since both technologies require compute and connectivity to function. Few companies exist in both spaces on the client and edge computing side since most specialize in one or the other. The few companies that currently play in both 5G and AI on the client side include Huawei, MediaTek and Qualcomm. Intel used to be part of this small cadre of companies, before it sold its 5G modem division to Apple. Now, for the most part, it turns to MediaTek for 5G modems. That said, Intel will still likely pair 5G and AI—just not to the same degree of integration as others. Additionally, I expect that Apple will begin to combine 5G and AI in the new iPhone 12—first with Qualcomm’s 5G modems, and then with its own home-grown modems sometime in the ballpark of 3-4 years from now.
5G and AI use cases
The potential use cases for 5G and AI together are boundless. While some are realistic in the short-term, many will likely require 5G and AI technology to mature to the point where application developers leverage both in their apps. The first and most obvious place where you can see AI and 5G working together, already, is in the smartphone. Last year, Qualcomm predicted there would be around 200 million 5G smartphones by the end of 2020, and over 450 million by the end of 2021. However, based on current subscriber numbers in China alone, we could see an even bigger market for 5G smartphones. Taking into consideration that each of these 5G phones already has a powerful SoC with AI capabilities inside, it is understandable that the industry sees it as lowest hanging fruit for 5G and AI. Currently, most of the 5G and AI capabilities in smartphones revolve around the camera, with filters for photos or AR applications like Snapchat. Apps like Google Photos also combine the power of 5G and AI cloud to improve object sorting and recognition in real time as you upload photos.
Another common, current use of these technologies is within voice assistants, which leverage 5G and AI to improve the quality and speed of speech-to-text and verbal search queries. It is expected that AI assistants will only get faster, more responsive, and more powerful with the low latency and high speed of 5G. We are already seeing enhanced video services that leverage 5G, such as Google’s Duo on the Pixel 5. The camera automatically defaults to HD, but adding AI to the mix enables background detection, and, in the future, noise suppression.
There are some privacy concerns with AI, which means many workloads may need to be handled on-device versus in the cloud. Companies will likely end up either splitting AI workloads between the two or keeping most of them on-device. Security follows many of the similar principles of privacy. That said, there are some security applications of AI that can be used to monitor a connection and block suspicious activity. These applications will need regular updating to improve the models, but they show promise.
5G and AI accelerating new industries
XR is one of the most mentioned ‘future’ use cases for both 5G and AI, and when you consider how much both improve the XR experience it makes complete sense. First, you have the streaming angle—the ability to stream high-quality content is a 5G use case on its own. When you add in AI-enhanced eye-tracking, though, you can reduce the rendering workload while also lowering the streaming bandwidth. This can improve performance, battery life while also maintaining overall perceived quality. Combining on-device AI-accelerated hand tracking with 5G enables a controller-free high-quality experience that leverages cloud-rendering. This would eliminate the need for a power cable, Wi-Fi connection or controllers, enabling applications outside of gaming, like retail. Retailers can leverage both hand and eye tracking to gain customer insights that help personalize next-gen retail experiences. These experiences can also leverage already existing models and renderings of products, as small as a shoe or as big as a building and display them in near photo-realistic quality. Headsets like the nReal Light and Oculus Quest, with Qualcomm’s XR2, are great windows into 5G and AI’s potential for XR. Companies like XRSpace, with its Manova headset, are already using 5G and AI to create new virtual worlds that allow for more natural human interactions.
Transportation is another opportunity for the pairing of 5G with AI. Consider the countless applications that we are already seeing. The two leading technologies that leverage 5G and AI in transportation at the moment are C-V2X and autonomous driving. C-V2X is an umbrella term that includes both vehicle-to-vehicle and vehicle-to-infrastructure communications—technology that leverages 5G and AI to make vehicles more aware of their surroundings including pedestrians to improve public safety. Leveraging 5G and AI to improve a driver or vehicle’s awareness beyond the vehicle’s internal sensors and utilizing roadside infrastructure can extend these capabilities to anticipate potential problems. C-V2X keeps vehicles abreast, in real time, of other vehicles and the traffic conditions, improving road and public safety and enabling faster and more efficient travel to (hopefully) avoid traffic jams and improve fuel efficiency and reduce emissions. To unlock the full potential of this technology, the corresponding roadside infrastructure, such as smart intersections or connected highways, will also have to grow, and improve.
Second, and related, is the autonomous vehicles sector. While autonomous vehicles are not quite here yet, they will leverage AI to make thousands of split-second driving decisions every day. To keep these vehicles aware of their surroundings and improve the on-device decision-making process, new data needs to be both downloaded and uploaded by the vehicles. While no network can handle the terabytes of data generated by an autonomous vehicle every day, AI will have a role in deciding which data gets uploaded to the cloud over 5G. Platforms like Qualcomm’s Snapdragon Ride which combine the SoC, accelerator and ADAS autonomy stack are poised to take advantage of this model in future autonomous vehicles. 5G connectivity also presents opportunities for smart assistants in vehicles, as do the increasingly powerful processors we are seeing. These AI-enabled chips perform fast natural language processing to execute navigation queries or other in-vehicle commands.
Last but certainly not least is the field of robotics. Naturally, many robots already have a considerable amount of AI, but adding 5G connectivity with platforms like Qualcomm’s RB5 gives robots a lot more freedom and autonomy than they have had in the past. While many people are familiar with private 5G networks for factory automation, many companies are also looking at using 5G to control and sync AI-powered robots with Ultra-Reliable Low-Latency Communications (URLLC), part of the 5G standard. This gives factory owners the ability to retool their factories quickly and easily without worrying about where they do or do not have connectivity. 5G’s low-latency and reliable connection should also bring about smarter and more capable AI-powered drones that can take over otherwise dangerous jobs like inspecting power lines or wind turbines. 5G connectivity enables the control and capture of live video feeds and recordings, while the drone uses AI to navigate around obstacles safely and determine whether something needs fixing or not.