Making Artificial Intelligence real
ASSA ABLOY started thinking about how to use Artificial Intelligence around three years ago. Since then, products have already been developed that use basic machine learning. And as the organization harvests more data, the possibilities to use Artificial Intelligence are growing and growing.
“Artificial intelligence starts with the data,” says Kenneth Pernyer, concept innovation manager for artificial Intelligence in the pre-product innovation department of ASSA ABLOY’s Shared Technologies unit. “And we have a lot of data. We often in fact refer to our work as data driven innovation.”
Kenneth and his team are completely dedicated to researching the use of Artificial Intelligence (AI). “We do early (Horizon-3) research mixed up with “here-and-now” possibilities” he says. “So the work we do now could be used in a product in a year or two or it could be that the experimentation is simply of great value in itself.”
We have a vision in which autonomous doors can understand their own service needs.
Potential applications of AI
One example of the products they are looking at now is smart doors. “We have a vision in which autonomous doors can understand their own service needs and fine-tune themselves using self-supervised learning,” he explains.
“And while on this journey we have found low-hanging fruit such as machine learning algorithms that tell us if a malfunction has occurred because a door has been hit by a vehicle.” Kenneth says. “This type of information would be great for something like an insurance claim.
“We are also looking at pre-break in attempts through the use of sensors on windows. We can now tell the differences between burglars using crowbars, children kicking balls against windows and the effect of a hail-storm. Recognizing those patterns does not necessarily require advanced AI, but it does require decent machine learning algorithms.”
Machine learning, which involves the automatic performance of tasks based on the interpretation of patterns of algorithms, is one of many subsets of AI.
“When it comes to machine learning you can use statistical methods that work fairly well with some data,” Kenneth explains. “But things like self-driving cars use variations of deep neural networks that require an enormous amount of data.
The more data you have, the more layers your networks can have and so you get better performance.
“We don’t have anything like that. Self-driving cars require networks that are over a hundred layers deep, whereas our current projects are around 10-20 layers deep. We are not Facebook or Google when it comes to the amount of data we have, but we do have some projects that have been online for a while with which we have been gathering quite a lot of data. And the more data you have, the more layers your networks can have and so you get better performance.”
Identifying AI projects
The projects that Kenneth and his team decide to develop are determined by a pre product innovation process. “We have this gut feeling that there must be something to learn from the data,” he says. “Then we start the machine learning project around it. After a while we will realize whether we have enough data, or that the data we have isn’t good enough. Then we can say that this is a good AI project but we need better or more data.
More and more ASSA ABLOY brands will ask for our advice and solutions based on AI
“Data is essentially an innovation tool,” he adds. “And more data brings more AI business opportunities which brings more data and so the virtuous cycle starts.”
ASSA ABLOY’s development of the use of AI in its products is closely linked to its evolution into the world of digitalization. “I think digitalization in general takes time,” says Kenneth. “But the important thing is that the process has begun. And as more people across the Group see the opportunities, more and more ASSA ABLOY brands will ask for our advice and solutions based on AI.”
Kenneth believes that the use of AI will be a gradual process with quantum leaps coming in a few years. “Those will be when we enhance our business models and build even more solutions around the doors in our homes and commercial systems that really benefit from the smartness of the solutions that we build based on lots of data. We don’t have this now but we will in a few years.”
Text: Danny Chapman