AIOT: Connecting AI To The Real World
Connected Ddevices Increase AI's Relevance
Analysts loosely define the Artificial Intelligence of Things (AIoT) as the convergence of AI and IoT — using AI to make IoT devices smarter and more autonomous. But that's a device-centric, "little data" definition. From a strategic, "big data" viewpoint, AIoT is the connection between machine intelligence and the physical world. The big data used for AI training and inference begins as little data captured by devices at the edge of the network that interact with things and people. These IoT sensors and human input devices are sources of truth that make AI relevant and valuable. In other words, connected devices comprise the nervous system connecting AI to our world. AIoT makes AI real and relevant.
AIOT and the value of device connectivity
AI creates insatiable demands for trustworthy, real-world data to drive training and inference. Thus, rapid AI growth requires enormous amounts of accurate data about our world, and this dependency fundamentally changes the economics of device connectivity. Instead of valuing a connected device in terms of the intrinsic value of its functionality, like a thermostat measuring temperature and controlling an HVAC unit, AI expands the device value proposition to include contributions to higher-level systems such as energy management.
total_device_value = functional_value + AI_contribution_value - device_cost
AI-based ecosystems extract more information from connected sensors and controls, potentially increasing the value of each participating device significantly beyond its base functionality. However, the value of each device's AI contribution varies by use case and solution scale, so AI isn't a blank check for making expensive devices. Functional value still sets baseline customer cost expectations, so the key to developing profitable AIoT devices is adding AI capabilities and ecosystem connectivity as efficiently as possible — and cost-effective connectivity is the main focus of our paper (see more information below).
For more information on specific AI use cases for IoT devices, please read the third paper in the Moor Insights & Strategy (MI&S) NXP Matter series, "Matter for CE Product Manufacturers".
Smart home AI examples include energy management, HVAC optimization, home security, safety, health and wellness, and aging-in-place. These advanced, AI-driven applications need situational context — the ability to perceive, understand, and respond to complex situations in real time. Therefore, AI applications must connect with multiple home systems such as lights, doors, windows, cameras, security sensors, HVAC, appliances, AV, plumbing, irrigation, pools, cars, and energy sources. Universal connectivity requires standardizing two things — device networks and an application layer. Let's use examples from the consumer electronics industry to see how this works.
AIOT Redefines Device Connectivity
AI's explosive growth motivates connected device manufacturers to meet rapidly increasing demand for IoT products that provide real-world inputs to AI-based ecosystems. From a connectivity perspective, this requires
(1) standard, widely available IP-based networks and
(2) a software application layer that provides direct, secure, multivendor communication between devices and applications.
This new definition of device connectivity combines message delivery with message content within a vertical domain. Standardizing connectivity accelerates product development, simplifies device installation, and reduces total product cost. Standardizing message content further increases device value by enabling useful connections with multiple applications and ecosystems.
The following two sections explore the practicality of building mass-market AIoT products using a surprisingly small number of IP-bearing networks and a domain-specific application layer.