What is the relationship between AI and IoT?

The impact of social AI on the Internet of Things

Will Cappelli
© Shutterstock / metamorworks  

With IoT, you are giving devices, vehicles, buildings the ability to host algorithms and perform functions which can only be driven by software. Therefore, AI software is a must when you need to handle the complexity of the Internet of Things. In this article, Will Cappelli explores two major intersections of AI and IoT.

There are two fundamental aspects which define the relationship and overlap between AI and the Internet of Things, and they’re quite different. The first is reasonably obvious. The Internet of Things enables the smart automation of many objects. With IoT, you are giving devices, vehicles, buildings the ability to host algorithms and perform functions which can only be driven by software. When you examine the tasks that these objects need to perform, the tasks frequently involve processes that are similar to the cognitive process.

What do I mean by this? Well, firstly you need to observe to understand what is changing in a complex environment, then there is an analysis of what has been observed, followed by the construction of an action plan to respond to what has been analyzed and observed, and finally, the loop closes with the execution of the action plan. It’s a four-step process which is very similar to how the brain works. In this scenario, the type of software you deploy will be AI. That’s the first overlap, the Internet of Things turns a lot of the physical world into robots and you need AI to drive robots.

Secondly, let’s look at the Internet of Things with an AIOps hat on. When you look at segments and domains of the Internet of Things what is very characteristic it that it is a highly complex interaction of individual objects whose individual behaviors are extremely difficult to predict. Due to this complexity, these environments are virtually unmanageable. If something goes wrong it’s hard for the human eye to identity what has gone wrong and consequently fix it.

In essence, what you need is a cognitive prosthetic on the human manager so the team can use AI to cope with the complexity of the environment. In addition, in an IoT scenario timeframes shrink, leaving the individual with microseconds to respond. The acceleration of state changes means that our minds struggle to comprehend what’s going on. Therefore, AI software is a must to handle the complexity of the Internet of Things. The software enables us to manage data at speed and scale, without the software humans fail. For me, these are the two major intersections of AI and the Internet of Things.

Social AI and IoT

I believe the Internet of Things will have a massive influence on the evolution of AI and I think the changes will be quite significant. When you examine AI that has been commercialized to date, the algorithms deployed are largely single agent, almost first-person algorithms. The algorithms are designed to see, analyse and act. There is interaction, but all of the intelligence takes place independently. Now, let’s look at other algorithms, for example those designed for a marketplace or trading floor. Here, prices get set as a result of many micro interactions between individual traders or reactions to issues in the supply chain or currency fluctuations. These interacting set of individual occurrences result in a price calculation for a wide variety of goods. This type of algorithm is what I like to refer to as social AI.

How is the world going to really benefit from the Internet of Things? It’s my view that it won’t be as a result of all these little robots interacting with one another individually, it will be the result of the planned, structured algorithmic outcome of that interaction. For example, look at self-driving cars, the algorithm is not just focussed on the individual car, it is also looking to optimize the distribution of traffic in a congested situation. So here we are interested in the social AI algorithm that is not just the sum of all these individual actors.

But how do you impose top down constraints i.e. mechanism design, to ensure you to get to the outcome you’re looking for? In the case of a market, you want the market to ensure that in a situation of scarcity goods are allocated to those who will provide the most value. The Internet of Things is a market of interacting individuals and forces, so what I think you will see is the forced commercialization of algorithmic game theory, mechanism design and social AI. Therefore, the next big thing in AI will be driven by the requirements of the Internet of Things which will culminate in a shift from individualistic AI to social AI.

The merging of AI and IoT

Undoubtedly, the Internet of Things and AI are seen as separate from one another. However, these disciplines will increasingly be seen as two sides of the same coin. This isn’t a new concept. Rather than thinking of the Internet of Things we will start thinking of a system of distributed, smart agents. What will we call this, who knows…but we are moving to a point where the default object won’t be a passive, dumb, software-driven device. It will be a combination of hardware and software that is making decisions within the context of other decision-making agents. That will be the real shift – moving from a bunch of devices connected to the Internet, to be a collection of interacting smart agents.

When we start looking at the world in terms of an assembly of smart agents, we will find human beings interacting in this community, some of us will be flesh and blood, other players will be digital, and very often the individual participants won’t have a clue, or won’t care, if the interaction is with a robot or human. In five years, we won’t be talking about the Internet of Things, we will be talking about a new buzzword that is describing smart agents interacting in this digital community.

Enterprise AI

To date, the obvious investment in IoT has focussed on the consumer, whether it be smart buildings, smart vehicles, smart phones or smart homes, as companies look to satiate consumer demands. As time moves on, this investment will become more enterprise and B2B focussed.

In the future, a lot of interactions between businesses will look a lot like a trading floor. When two algorithms are buying and selling, they are acting like business people, and I believe this will spread, not only in terms of day to day transactions, but also into the automation of the supply chain. In many respects the way in which markets work, is the model for how AI and the Internet of Things will converge and evolve. The structure of the economy is ready made to be mechanized – all we’re doing is automating a process whose structure is already an algorithm, it’s just not coded. Let’s call this the algorithmization of the economy.

SEE ALSO: Say hello to a new AIOps platform: Wave of the future or just another drop in the ocean?

The evolution of AIOps

To manage the complexity of the Internet of Things system you need AIOps, it’s not an option – the reality is that the Internet of Things cannot be corrected without AIOps.  Many people orchestrating Internet of Things projects will be aware that AI is needed if that project is to be successful. However, they will not be aware of the need for management disciplines to ensure the Internet of Things is delivering what it should.

AI is applied to complex environments, but we haven’t thought consciously of what it means to apply management to a community of smart agents. How do you do AIOps, or ops management, when Skynet becomes self-aware? When we reach that singularity, and I don’t think we’re far away from it, then we should expect a reformation of management disciplines.

I can imagine management disciplines of the future will come to resemble macroeconomic policy management techniques which are used by organizations like the Bank of England or the Federal Reserve. This is how I see AIOps evolving in the future.

Will Cappelli
Will studied math and philosophy at university, has been involved in the IT industry for over 30 years, and for most of his professional life has focused on both AI and IT operations management technology and practises. As an analyst at Gartner he is widely credited for having been the first to define the AIOps market and is CTO EMEA and Global VP of Product Strategy at Moogsoft. In his spare time, he dabbles in ancient languages.

Inline Feedbacks
View all comments