Bringing the AIoT to Life: Pairing Machine Learning with Streaming Data

The coupling of Artificial Intelligence with the Internet of Things represents a natural progression for applications in both these spaces. Machine learning and AI deliver a number of benefits for the IoT which, because of the rapidity of the data it generates, assists AI in multiple ways.

For the IoT, tangible advantages of this combination include:

  • Increasingly Sophisticated Technologies: AI provides a surfeit of technologies including facial recognition, speech recognition, and others to enhance the capabilities of the IoT to include critical deployments like video camera security use cases.
  • Real-Time Analytics: The predictive analytics AI supplies are ideal for rapidly processing data in real-time for low latent decision making required for applications like autonomous vehicles.
  • Intelligent Edge Deployments: Edge use cases generate much more value with machine learning and AI to improve analytics in endpoint devices, requiring less data and decision-making in centralized cloud settings.

In turn, the IoT is the ideal milieu in which to train AI models because it produces an enormous quantity of data (which those models necessitate) at high velocities. These capabilities enable such models to continually be refreshed for the most informed predictions possible.

Model Training

One of the inhibitors to enterprise applications of machine learning and AI is the dearth of training data (and annotated training data, at that) in a specific user domain. Therefore, when organizations are able to access such data, the models of their “traditional AI relies on historical data,” explained Max Nirenberg, CRO of Commit USA. With numerous devices creating sensor data and streaming data, the IoT is acclaimed for its huge amounts of data that are continually produced with low latency. Consequently, “with more data in real-time happening all the time, this makes historical data less relevant,” Nirenberg noted. In fact, one of the more meaningful developments to impact contemporary data science is the ModelOps movement in which progressive organizations are able to train and deploy data simultaneously (at the edge, in some instances) via the IoT.

This technique and others are pivotal for time-sensitive use cases such as the one Nirenberg described in which an IoT device “has face recognition, and it also recognizes voices. It recognizes text, but it also recognizes when there are two objects next to each other that shouldn’t be there.” Such capabilities are essential for security monitoring in private and public places like airports, for example. “The device that I’m referencing is one that we created for the defense industry,” Nirenberg acknowledged. “It was definitely created from… think of like a terrorist component, but it has many applications. It doesn’t have to be just that.”

Predictive and Prescriptive Analytics

The reciprocal boons of the nature of the relationship of the IoT and AI are demonstrated in the fact that may IoT use cases either directly hinge on, or are considerably enriched by, what Nirenberg termed as the capability to “do analysis based on data at scale.” This quality, of course, is implicit to almost any deployment of machine learning and Artificial Intelligence, which naturally reinforces the compatibility of these technologies with the IoT.

The motion towards developing smart cities and smart buildings avails itself of this potential in a couple ways, particularly when considered in light of the ongoing public health crisis. In a smart building, for instance, it’s much more than a matter of academic interest “to know who’s entered which rooms to make sure that the appropriate people were allowed in and the inappropriate people were kept out,” Nirenberg commented. “With COVID, I may want to know where there’s high levels of foot traffic, and I need that data quick.”

Made for Each Other

AI’s timely processing of data for this use case and others–whether involving object detection or some other form of analysis–can provide this intelligence to derive low latent action. The real-time nature of this action, as well as the nearly instantaneous analysis from whence it springs, is the cardinal value proposition of the AIoT. When used in tandem with the IoT, AI’s low latent predictive analytics becomes imbued with the ability to monitor quotidian developments that impact the quality of life for consumers, the enterprise, and even the public sector—or society in general.

“It’s being able to take any kind of regular everyday thing that you do and obtain information from it,” Nirenberg reflected about the IoT. “It lets you get information so you can then make decisions like, oh my god, send people to that place right now.”

Featured Image: NeedPix 

About The Author

Scroll to Top
Share via
Copy link
Powered by Social Snap