AI into Things.

AI (Artificial Intelligence)-enabled machines are programmed to “think” like humans and imitate their actions. These AI machines can also perform human-like capabilities such as learning, problem-solving, or decision-making. All these capabilities allow AI to be the right choice for applications such as natural language processing, speech recognition, and machine vision.

AI can add value to any machine, and that includes IoT (Internet of Things) devices. AI can improve the capabilities of IoT through Machine Learning (ML) capabilities and enhanced decision-making processes. But AI, in the same case, can also benefit from this “AI-IoT relationship.”

The convergence of these two technologies— AI and IoT, can create more intelligent machines that simulate smart human behaviors and make decisions with little to no human intervention.

AI by itself can’t generate data and much less transport it. IoT can add value to AI technology by adding connectivity and data exchange capabilities. IoT devices capture a massive amount of data from multiple sources. But therein lies another challenge; this massive amount of rich-context IoT data can be beneficial until it starts taking too much storage, or starts becoming difficult to collect, transport, and analyze.

What is AIoT, and how did it start?

As highlighted in the section before, AI and IoT can complement each other. IoT, for instance, deals with devices that perform an action, such as data collection, and also devices with internet connectivity. At the same time, AI makes these devices learn from their generated data and experiences.

So, what is AIoT? AIoT is the concept that defines the relationship between AI and IoT technologies. In simple words, AIoT brings AI to IoT devices.

To help explain this concept further and how it came to exist, it is worth mentioning “edge computing,” another trending topic. Edge computing is a type of “computing power” that moves far away from a centralized computing architecture, such as cloud computing. It is a combination of hardware, software, and practices that aim to decentralize computing capabilities and demands and thus bring intelligence to the edge— closer to where data is generated, where devices such as mobiles, cameras, sensors, or IoT are. Edge computing also opened the doors for other related technologies, such as edge AI. The concept of edge AI is, in layman’s terms: “any technology or practice that aims to bring AI closer to the edge.” Edge AI appliances, for instance, could help monitor health, analyze video in real-time, automate robots, manage traffic, and many more.

Now that you know about edge computing and edge AI, it can be easier to understand AIoT. So, AIoT is simply an edge device with Internet connectivity (IoT) running some sort of AI capabilities.

Benefits of AIoT Devices.

To depict the benefits and challenges of AIoT, let’s take an autonomous self-driving vehicle as an example. This vehicle has sensors, cameras, and other IoT devices that collect data. The vehicle needs to process this collected data in real-time— or else there is a risk of accidents. Processing this data or applying AI at the edge (edge AI on the vehicle itself) reduces the need to move the data to a remote server. In addition, when the vehicle is traveling through areas with low (or no) Internet access or is going fast, the edge processing at the edge could be really beneficial.

An AIoT device is an IoT device with the right hardware and software capabilities to run edge AI. One of the benefits of an AIoT device is that it can be independent of remote cloud-based computers running AI workloads. But if it needs to send data and be connected to the Internet (as is the case for IoT devices), then the edge capabilities remove any transport delays or latencies in the process.

The Summary of AIoT Benefits:

  • AIoT aims to perform better IoT operations.
  • AIoT attempts to get more value out of IoT-generated data.
  • AIoT also aims to improve IoT’s data analytics and its management.
  • With AIoT, AI helps the IoT device to use the collected data to better understand, analyze, learn, and make quick and important decisions.

The Challenge with AIoT.

Although AI has been around for a while, in practical use, AI has been a technology only accessible to enterprises with data centers and robust high-speed computers. AI needs powerful hardware to handle its workloads and process large amounts of data. Enterprises can afford that, but not the majority.

Fortunately, many cloud providers such as AWS, Azure, and GCP provide robust resources (as a Service) so that the majority can rent these resources and run AI. This is probably the right approach for an AIoT device; to access cloud-based AI services via its Internet connectivity.

But although renting resources from the cloud has been an attainable solution, there are a few problems for AIoT devices:

  • IoT devices capture a massive amount of data from multiple sources. This massive amount of rich-context IoT data can be beneficial until it starts taking too much storage or starts becoming difficult to collect, transport, and analyze.
  • Network constraints and cloud costs Renting cloud-based hardware for AI tends to be more expensive in the long run. Plus, since you are outsourcing power (to the cloud), you’ll need robust network capacity to handle time-sensitive workloads and massive data transfers.

Don’t worry, we’ve got 5G!

One of the best innovations to AIoT is the incorporation of 5G; therein comes the concept of 5G AIoT. This next-gen mobile technology, along with AIoT, can help with all the challenges mentioned above.

5G in a nutshell: 5G is now a mainstream wireless technology available for mobile devices. 5G is different than its predecessors (3G and 4G) simply because it is designed with ubiquity in mind; it connects virtually everything, including machines, devices, and anything that supports it. 5G’s main features include its ultra-low latency, peak data speeds, reliability, availability, and network capacity. These features inevitably contribute to a better user experience and more connectivity.

Although such features are related to network (and not to computing as AI need), they significantly contribute to AIoT. They provide the adequate highway that a terminal needing AI (hosted in a remote server) needs. It allows AIoT devices to send massive amounts of data without any delay and latency.

Example of a 5G AIoT appliance.

Lanner’s EAI-I130 is an industrial-grade Edge AI appliance powered by NVIDIA® Jetson Xavier NX. Nvidia considers the Jeston Xavier, NX Series Modules as the world’s smallest Artificial Intelligence supercomputers. This edge appliance (EAI-I130) achieves up to 21 TOPS for AI performance. The AIoT appliance also supports 5G and Wifi6, making it a real 5G AIoT appliance. In addition, this appliance is also designed to work in challenging environments. It is IP40 compliant and works in temperatures ranging from -40°C to 70°C.


Next Steps.

For more information on the 5G AIoT appliance or the edge AI with 5G support appliance, please contact Lanner’s sales representative.

Featured Product


Industrial Grade AI Inference System For 5G Edge With NVIDIA® Jetson NX

CPU 6-core NVIDIA Carmel ARM®v8.2 64-bit CPU 2MB L2 + 4MB L3
Chipset N/A

Read more