You hear a lot about the Internet of Things (IoT). In fact, lately I've been talking about it a lot at least. Another topic that is taking up a lot of space is... not Netflix series, but rather artificial intelligence (AI).
There is a very strong cohesion between the two that gave the idea to some to create a new acronym, AIoT for Artificial Intelligence of Things. But what is this? Take a deep breath and let's look at it together.
Use of the data
In a recent article, I told you that IoT is actually the connectivity component of a digital solution. Indeed, you just need to put a SIM card, a network cable or connect an object, a machine, a car and voilà! You get an IoT device.
Is it that simple? Conceptually speaking... yes. However, there is unfortunately no magic that makes an object start using that connectivity link to send data. Also, where should that data go? It seemed so simple at first! Well, it is, but it requires some special skills.
For connected objects, it is usually enough to modify the code contained on the device in order to use the connectivity link and send data to the location of choice. Whether it's in the cloud, the Edge (now you know what that is since we talked about it in another article) or simply directly into your servers. Of course, in addition to knowing how to program, you need to understand how the communication link works. Another knowledge to add to the list.
Ultimately, the data is stored in a database. There are many solutions available to you, from relational databases (or not), to data lakes and so on. Once again, you have to know these options to be able to use them.
When you're lucky, you manage to find the Swiss army knife of the IoT: a developer who does embedded programming, who is good with communication protocols, who masters the cloud and who manages databases. If you find this rare pearl, tell him to call us... I'm kidding, but still, it's not that easy to find.
The kind of IoT possible
Now we have everything we need to do IoT! Actually no... it is possible to make only one type: IIoT... A new term! Yes, as IoT can easily be applied to many things, we just have to add a letter and it becomes a solution. IIoT stands for Industrial IoT, which is in a way a component of Industry 4.0.
As I said, if we want to do IIoT, we will need someone who knows about machines (PLCs) and their protocols (Modbus, OPC UA, etc.). More acronyms! In short, an automation engineer usually does the trick. But this person is rarely the same as for the other expertises we named earlier.
AI + IoT: The perfect match
Let's get back to AIoT. The result of IoT is that we accumulate data from an object, machine or whatever. This opens up a plethora of possibilities! We can do visualization using dashboards and business intelligence (BI). When we have enough data, it is even possible to do artificial intelligence with it all... hence AIoT!
Indeed, AI is often the logical continuation of IoT. The wealth of information extracted by the IoT easily feeds artificial intelligence models to solve all sorts of problems. Whether it's predictive maintenance, classification (recognizing a dog from a cat or something more advanced like recognizing the type of disease in a plant) or simply trend analysis, the combination of IoT with AI makes Cupid completely jealous for not having thought of such a perfect match.
The experts behind AIoT
Now, let's add to our already very complex toolbox. Not everyone can become an AI expert. It takes specialists who have a very good grasp of the mathematical models (mainly probabilities) behind the various models.
But that's not all. It's not enough to just take the data and shuffle it into an AI model. There is a lot of work to be done on the data before it is possible to get something out of the AI. So we're adding another body of work, which is data scientists.
Who are they? What do they do? (Sounds a bit like Goldorak!) But no, they are not robots. They are data scientists who can do the magic to make up for the fact that a sensor didn't send all the expected data or that there was interference that caused some completely wrong data to be sent.
So we have to clean up the mess, because if we don't send data to the AI model in the expected format, the result will be as bad as the quality of the data we provide. So we have to do regressions, filtering or other data cleaning techniques.
Let's recap. So we have an automation specialist, an embedded systems programmer, a connectivity specialist, a cloud specialist, a database administrator, a developer to do the application portion, an integrator to move that data around to create alerts or feed the AI portion, a data scientist and an AI model expert. That's a lot of people.
But that's not all! There are other factors to consider in all of this. You have to operationalize the model learning (ML-Ops), manage the deployment of the solution (Dev-Ops) and also, a portion that is becoming more and more important, manage the cybersecurity of the solution (Sec-Ops). With everyone invited to this development, it takes a strong project manager and an architect to ensure that all the pieces fit together. We mix it all up and it gives us a great digital transformation solution.
As a customer, you are in the best position to know what the data means, what is normal operation for your machine and more. It is indeed very difficult for someone who does not know the field to see that a correlation is completely wrong or does not make sense, much like this example
It might be easy to get excited about having found the life-saving relationship by getting Nicolas Cage to stop making movies, but in the end, I have big doubts that it will change anything. The same thing can happen when technical experts don't know the ins and outs of a business.
In short, AIoT is a logical continuation that leads to a global intelligent solution (Smart Building, Smart City, etc.). It's a bit like a symphony orchestra. You are the maestro and you lead experts in their instruments. The only difference is that we're not talking about violins, trumpets or anything else, but rather C, LTE/5G, Azure, MongoDB, TensorFlow and more.
The result is a work of art that will help you achieve a goal: save time and money, increase quality and so on.
Do you have an AIoT project in mind? We want to know more! Contact us now to discuss it.