IoT systems are good at one thing:
asking what is happening in the physical world?
A sensor can tell you what the temperature is. A connected asset can tell you where it is. Equipment can tell you how it is operating. A vehicle can tell you where it is going.
But when an organization has thousands or even millions of datapoints, a different question becomes more interesting:
what does it all mean?
This is where Artificial Intelligence and the Internet of Things come together, in an approach sometimes called AIoT.
IoT Provides the Data Layer
On a fundamental level, IoT is about connecting physical objects to digital systems.
An industrial IoT system can include information from:
Equipment or machinery
Vehicles
Physical assets
Inventory
Environmental sensors
Connected tools
Industrial processes
The information from these sources can give an organization greater insight into its physical environment.
But sensor data is generally not enough to make a complex operational decision.
For example, a sensor value change tells you that something happened; figuring out what something might be requires additional analysis.
This is where the power of AI can be productively employed.
AI Turns Data Into Context
AI and machine learning can ingest massive amounts of IoT data, and extract meaning, identify patterns, detect anomalies, enable forecasting, and more.
The rough structure looks like this:
IoT → Data → AI/ML → Insight → Decision
IoT provides observations; AI helps make sense of those observations, and extract meaning from them.
This, in turn, can be used to make an operational decision.
Notice that we're talking about something more interesting than "just" IoT.
AIoT is not about collecting data; it is about making connected data more valuable.
A Practical AIoT Architecture
An AIoT architecture could be conceptualized as a pipeline:
Physical Asset
↓
Sensors
↓
Connectivity
↓
Data Pipeline
↓
AI / ML
↓
Decision
↓
Physical Action
Each layer has its own set of engineering implications, which I will discuss shortly.
- Physical Asset
The system begins with a physical object - a machine, vehicle, tool, inventory item, or some other process.
- Sensors
The value of the physical asset is captured by some sort of sensor, which in turn produces some sort of usable data.
Without data, nothing else is possible.
- Connectivity
Connectivity is required to get the data from the physical world into the software environment.
Depending on the use case, a variety of networking and communication technologies may be used.
- Data Pipeline
Sensor data needs to be collated, processed, transformed, and aggregated for consumption by the AI/ML layer.
This is why the data pipeline layer is critically important, as the volume and variety of data increases.
- AI / ML
Artificial Intelligence and Machine Learning models can process the data, identify patterns, detect outliers, and more.
- Decision
Unless there is a decision to be made, the value of the AI/ML layer is extremely limited.
The decision might take a variety of forms: it could be related to maintenance, or to the health of an asset or a piece of inventory, or to operations or logistics decisions.
- Physical Action
Finally, the system makes some sort of impact on the physical world.
This is what makes AIoT fundamentally different from just an AI/ML system.
Industrial AIoT Is a Systems Problem
One of the common missteps with AIoT is to think of it as an AI problem with an IoT flavor.
In practice, however, it is crucial to think of the entire system.
It does not matter how good your model is if your sensors are not accurate. It does not matter how good your sensors are if the connection is unreliable. It does not matter how good your data pipeline is if the insights it produces are not valuable to the people who need to make decisions based on them.
There is no single silver bullet.
AIoT requires attention to data quality, to sensor accuracy, to connection reliability, to the data pipeline, to the model itself, and to the systems into which it needs to be integrated.
From Industrial Problems to Venture Opportunities
There is one other interesting aspect of AIoT.
Industrial systems tend to have a wealth of interesting problems related to assets, equipment, inventory, safety, logistics, and more.
Some of these problems can lend themselves to the creation of new products or even new ventures.
Venture studios can use the wealth of opportunities in the industrial space to identify promising problems, explore the available technological solutions, and determine whether a viable venture exists at the intersection of an interesting problem and a plausible solution.
This is the space in which Aperture Venture Studio operates.
The Bigger Picture
The evolution of the IoT is not a straightforward one.
Sure, we are rapidly approaching a future in which everything is connected.
But the more interesting evolution is:
Connect → Collect → Understand → Decide → Act.
IoT provides the connection and the collection.
AI can provide the additional layer of understanding.
Software systems can make decisions based on that understanding.
And those decisions, in turn, can have an impact on the physical world.
This is the promise of AIoT: to connect the dots between connected physical objects and the decisions that need to be made about them.
And this is why the opportunity for developers and engineers is much bigger than just building sensors or just building AI/ML models.
The true opportunity lies in building the entire system that connects the two.
Top comments (0)