A special thank you to all the attendees, vendors, and partners who have supported AT&T SHAPE. There are currently no planned dates for future SHAPE events, this site will be retired in 2021.
×
host_shape-https://shape.att.com
host_ddp-https://developer.att.com
env_shape-prod
env_ddp-prod
Published: Mar 11, 2020
Updated: Mar 11, 2020

The Powerful Combination of Machine Learning and 5G Networks

Author: Ed Schmit

Ed SchmitAuthor: Ed Schmit, AVP Product Marketing Management, AT&T Developer Program

Ed tracks new technologies for the AT&T Developer Program. His specialties include network technologies, technology enablement, and strategic marketing.

Every market and industry seems to be investing big money in artificial intelligence (AI) and machine learning (ML). Cloud software, Internet of Things (IoT), fintech, big data, no matter where you look, AI is there.

In most markets, however, we’ve only just scratched the surface of what AI can do for us. We’re only using the most basic functions, namely automation, which performs tasks faster, more accurately, and at a much lower cost, than any human could. Some technology will be so dependent on AI to function that we’ll be finally using it to its full potential. This is true for one of the hottest emerging technologies out there, 5G.

One of the fundamental characteristics of 5G is the ability to predict activity across the network and manage it seamlessly. Machine learning is ideally suited to working in 5G networks because it requires massive amounts of data to predict activity accurately. — an ideal situation for 5G, since it can transmit higher volumes of data faster than current networks.

So, what makes AI and ML such a big part of 5G? Let’s dive in.

Machine Learning Makes It Easier to Predict

Machine learning will be a necessary part of any 5G network since it’s more complicated than previous generation networks. It operates at higher frequencies, introduces more complex antenna/cell configurations, and employs more sophisticated connectivity mechanisms like beamforming. The multiple-input-multiple-output (MIMO) antennas used by 5G networks can handle multiple data “conversations” over the same data signal simultaneously.

All of this means that more data can be transmitted in any direction on the network without negatively impacting the transmission of any other data on the network. As more devices come on board a 5G network, it becomes impossible to handle all the traffic necessary without AI and machine learning. A 5G network will be able to analyze historical data patterns and conversation, allowing for more efficient transmission of data at all times. A fully functional 5G network will not happen without AI that can learn and make decisions by itself. These systems must be predictive and proactive on their own, so the data gets where it needs to be. Without it, future applications of 5G will be handcuffed, and we will be limited in the opportunities we can dream about.

AI Offers Network Optimization Opportunities

From a strict technology perspective, AI and machine learning offer several ways to boost network performance. Machine learning can help optimize overall network management and monitoring, drive efficient resource consumption, and enable customized network layering (or slicing, as it’s known) to give owners more control over their network usage.

Boosting Performance and Cutting Costs

By analyzing the usage of the system, machine ML and AI systems can determine and optimize device mobility patterns and quality of service usage to better predict network usage and congestion at specific locations throughout the day. They can schedule traffic and allocate resources more efficiently, ultimately providing better network service to users with lower resource consumption.

Optimizing network management

Machine learning provides additional insights into the health of the network by providing other capabilities and functionality for fault, performance, and security management. IT pros can more easily monitor the network and are notified of issues before they become problems. For example, a failing network device is discovered more quickly based on the AI analysis of historical traffic patterns through it. System failures are swiftly bypassed as they’re identified before they can critically impact the network and are mitigated easily, something difficult to do in highly complex systems without AI help.

Network slicing

To limit the number of hardware devices needed to power a 5G network (which typically requires significantly more than a traditional 4G one), IT pros have introduced the idea of virtualization (or network slicing). Network slicing is the idea of using a single, shared physical network with multiple virtualized networks running on it; It’s the 5G equivalent of a virtualized server.

For example, companies could run one network slice for their low latency-use systems that need high reliability on the same device that’s running another network slice that handles employee communication traffic. A 5G-connected car could have one slice for autonomous driving and other mission-critical functions, and a separate slice for infotainment and in-car controls like the AC.

Setting up network slices on a 5G network is limited right now, however, since each slice must be manually configured. ML and AI will facilitate this setup, no matter how complex the individual system or requirements. They will ensure that all communication traffic is routed based on device needs and the appropriate configuration settings.

Moving Beyond the Network

As we continue to uncover the ways machine learning can help us use 5G-powered networks to their fullest potential, it will expand the ways we use technology. It will drive innovation in every industry, as business leaders discover the ways it can transform the way they do business.

From IoT to predictive medicine

We’ve got a lot of IoT devices around us today. As more devices are developed and installed where we live and work, they’ll generate more data. Data about human patterns of work and living that will be available for machine learning analysis that can help us work and live better. Far from the scary idea of the robots taking over, ML will help IoT devices work better, creating a better lifestyle for us — which is what they’re intended to do in the first place.

Technology in Healthcare

A few years ago, it was strange to see a doctor typing out their notes on a computer, but now it’s commonplace. We’re seeing more technology come into use in a variety of healthcare settings, making it easier for healthcare professionals to do their work, while also helping to educate patients better so they can be better advocates for their own care. Some platforms help doctors provide rapid and accurate diagnoses, ones that streamline treatment plans, and others that transmit data efficiently between healthcare practitioners. Machine learning and health-related IoT devices will help make all of these interconnected systems work more efficiently and rapidly, an important thing in healthcare.

Healthcare generates a lot of data, which is something AI needs for its learning process. Much of healthcare data uses a lot of bandwidth, whether because it’s high-resolution images or just the sheer volume of information required for each patient (test results, patient histories, etc.). Trying to move all this data through traditional networks just isn’t feasible since it must be done through reliable networks with high capacity. Switching to a 5G network lets healthcare institutions deploy machine learning tools on a fast network that ensures the system will be highly available and extremely efficient.

5G-powered Predictive Medicine

Once healthcare starts using 5G and AI to manage their existing systems, they’ll turn to new ways of caring for people. One area where IoT, 5G, and AI can be especially useful is in medical studies. No matter the topic being investigated, medical studies strive to track as many patients as possible over the longest period of time possible to see the effects of illnesses, treatments, lifestyles, location, and more have on them. Instead of searching for patients that meet specific criteria and then signing them up to the study, they’ll be able to use smart health tracker data.

Most of us already wear basic health trackers, so it only makes sense that in the future, we’ll have more sophisticated ones that track even more parameters. The information from our trackers will be sent via a 5G network to an appropriately secured cloud location where it can be aggregated and processed. Adding AI-powered algorithms could ensure new insights would be derived from the same health data, providing people with better ways to stay healthy and live longer lives.

Challenges to a 5G-powered AI Network

The 5G rollout is only just beginning, so understandably, the industry hasn’t really explored what 5G can do with machine learning. There are several obstacles to implementing 5G at scale so that it’s usable by everyone.

Dealing with Interference

For starters, a 5G signals are more prone to interference as they pass through physical objects. To combat this, 5G networks will be powered by smaller “stations” spread throughout locations at closer intervals than the 4G network towers we’re used to seeing now.

Dealing with Latency

Another obstacle to 5G networks running AI software is centered around where to run the AI software. These systems are very demanding in terms of computing power and low latency needs, because without those, they’re not reliable. Likewise, a 5G network: it absolutely requires low latency, otherwise, it’s not useful. If the data the AI needs is stored in a cloud system far away from the AI software, even with a 5G network powering it, there will still be too much of a delay for the AI’s work to be as useful as it could be.

Simply put, latency increases with distance and congestion of network systems. Depending on the system using the 5G network, the latency could become a critical problem. For example, autonomous vehicles will not work with high latency systems because it must recognize objects like pedestrians, in real-time. Microsecond delays in response times could have catastrophic implications for both passengers and those outside the vehicle.

Solving Latency

A solution to this issue would be to include edge computing systems within the network, so it can move the data and/or computing power needed by the AI closer to the AI that’s doing the work. The 5G network can make all of this happen more efficiently, allowing the AI to do its job efficiently too. AI software can help make the entire network more predictive, directing traffic to the appropriate device or system as needed, so it’s fully optimized to handle whatever data is on it.

Final Thoughts

The rollout of 5G will enable a new era of opportunity for everyone. The ability to collect and transmit data from anywhere in the world to anywhere else in the blink of an eye will revolutionize the way we think about and use technology. It will unlock the creative minds of technology professionals as they think of new and innovative ways to improve our businesses and lives.

Adding AI and machine learning to the 5G mix will make things even more interesting. Smart devices will be able to share information with systems designed to help optimize our lives. It started with the smart thermostat that “learns” our daily schedule at home and changes the temperature of our living quarters as we come and go. It’s progressing to the integration of IoT devices in a business setting that’s helping companies control access to physical locations, monitor IT systems for intrusions and faults, and AI software that’s optimizing network traffic to enable edge computing and help use process information more efficiently.

With the right combination of AI on a 5G network, we’ll enjoy unprecedented levels of performance and automation on our mobile networks. It’ll drive innovation in all aspects of life, changing the way we live and work for years to come.

 


Share this post