A great way to think about this is to understand how AI and IoT are fundamentally different technologies.
A helpful way to think of it is:
Iot connects the physical world to a digital space, whereas AI supports interpretation, pattern recognition, and decision making.
When it comes to developers, it is interesting to understand their implications when both are composed as a single system.
What is IoT? What does IoT do?
Internet of Things (IoT) is focused on the physical, sensing aspect of things.
An IoT system can feature:
Sensors
Embedded devices,
machines or equipment
network connectivity
gateways
edge or cloud infrastructure
data platforms
etc.
Imagine a physical machine with sensors that read out temperature, vibration, pressure, and other metrics. A simplified view of such a system could be:
Physical machine
Sensors
Embedded device / Gateway
Network connectivity
Data platform
This gives software a window into the physical world.
However, it is only one part of the story.
Where does AI come in?
AI works on the data produced by IoT systems.
Whereas IoT will be able to provide information about the physical world, AI can analyze data in order to find patterns, irregularities, predictions, classifications, and more.
Imagine a machine that typically vibrates within a normal range of operation. IoT gives a view of the machine's vibration. AI can recognize patterns in the data and identify if the machine falls outside of the normal range.
IoT asks "what is happening in the physical world", whereas AI can support analyzing the data and finding patterns worth exploring.
IoT and AI are complementary
IoT and AI are both fundamentally different technologies, yet they can work together.
IoT offers connections to the physical world, whereas AI provides algorithms for analysis.
However, putting the two together may not automatically lead to an "intelligent system".
AI is only as good as the data that is provided to it, and IoT data may vary in quality.
On the other hand, merely collecting data about the physical world does not always yield information that is immediately valuable for analysis.
As a result, many factors have to be taken into consideration in order for an AIoT system to be successful.
Developers may want to account for:
sensor accuracy
data ingestion
connectivity
latency
data quality
storage
model performance
security
reliability
human factors
etc.
What is AIoT?
As mentioned, when AI is added to IoT it forms an AIoT system, also referred to as Artificial Intelligence of Things.
The architecture can take multiple shapes depending on the use case.
Some systems process data closer to the physical equipment (edge AI), some route it to centralized cloud infrastructure (cloud AI), and some do a mix of the two (hybrid). It is important to account for the implications of each approach.
Connectivity, bandwidth, latency, processing power, security, and data privacy are important factors to consider when designing such a system.
What developers should consider
An AIoT system involves more than just an AI model.
A developer looking to build such a system could benefit from designing each step of an end-to-end system.
Sensors
Device / Gateway
Connectivity
Data ingestion
Storage / Processing
AI / ML
Application
Decision / Action
For every step there are implications that a developer should explore and understand.
For example, machine learning models are often designed with a certain set of assumptions, and may underperform when deployed in production due to differences in deployment environments such as embedded systems, edge devices, etc.
On the other hand, IoT systems rarely provide structured data, and often have to go through additional steps of preprocessing, filtering, normalization, and feature engineering prior to being fed into a machine learning model.
Additionally, the physical environment plays an important role in an AIoT system.
Unlike building a traditional software application, an AIoT system has to deal with limitations and reliabilities of hardware, connectivity, and sensors.
In short, developers need to account for both the data coming from IoT systems, as well as the limitations of the physical systems supporting these data.
From connected devices to intelligent systems
One of the main benefits of AIoT is the notion of connecting AI to the physical world. This allows for additional insights that were previously impossible.
AIoT can be applied to a wide variety of industrial applications, including machines, equipment, assets, plants, logistics, and more.
On that note, Aperture Venture Studio has more information about AIoT and Physical AI in industrial spaces.
The main takeaway is that IoT gives software access to the physical world, and AI helps software understand it. Developers can use this combination to build valuable systems that go beyond monitoring and provide new insights and intelligence.
Top comments (0)