The Amsterdam-based software company BrainCreators is betting on a future where no one does boring and repetitive work anymore. And no, the company is not working on a robot that will take over these tasks but on another technology that can automate these tasks: digital inspectors.
Now the term "digital inspectors" probably sounds a bit vague to you. Yet it is relatively simple to explain because it is essentially what the name suggests: a software program powered by artificial intelligence.
"For example, we make software under the name INSPECH that can recognize damage in the road and report it to the person who needs to know," explains Jasper Wognum, co-founder and CEO of BrainCreators. "We are now doing this, for example, for Unihorn, a subsidiary of Strukton. And with this, we make it possible to check a stretch of highway for damage within minutes which used to take weeks."
Artificial intelligence: learning software
The technology is based on machine learning. This means that you give a piece of software a gigantic amount of data and tell it how that data should be interpreted. Based on such a dataset, the software can learn to recognize new examples it sees.
So in the case of Unihorn, the first thing that had to be determined was what types of road surface damage the system needed to recognize. To do this, cars mounted with cameras went on the road and collected many examples of damages.
The collected data was added to BrainCreators' INSPECH program and the result was software that can now recognize a wide range of road damage, such as cracking and fraying (stone loss).
This is a very specific example of a product BrainCreators makes and delivers (the company works in many industries), but it wasn't always that specific. "When we started BrainCreators, we had an idea of what we wanted to do, but no idea in what market it would be useful," Wognum explains. "We started scanning different markets and ended up sticking to a few interesting ones."
Scanning where artificial intelligence is most effective
At first, animal breeding seemed like a promising use case. The software that Wognum and colleagues developed for this purpose was able to predict, based on genetic data how much fat the pigs from a particular breeding program would have between the bones without actually having to breed, of course.
"When we told this at a conference dealing with the subject, there was a deafening silence after the presentation. The geneticists in the room were clearly impressed. But after the silence, the comments came. It was all impossible and unfeasible what we were showing. In retrospect, you think it's only natural that there would be comments when you share this kind of information with geneticists who at that moment are seeing their jobs evaporate," laughs Wognum. "We were obviously too early. But it also turned out that this was not a scalable product. In fact, only two companies are working on it worldwide."
Scale is important because making a 'digital inspector' takes time and money. This is mainly due to a large amount of data that needs to be collected. "That's why we work with domain partners," Wognum explains.
"We find a partner with whom we are going to work with for five months to build the digital inspector. Once that's done, they can get started, but they may also deliver the product to industry peers or other parties who can make some use of it. For example, Unihorn provides its services based on BrainCreators' software to BAM."
Unburdening employees with AI software
Wognum has another example where the process was more manageable. At Tata Steel in IJmuiden, BrainCreators worked on a system to recognize errors in products on conveyor belts, something that was done by a team of people previously. As a result, there was naturally some resistance because employees thought that automation using smart software would be a threat to their own jobs.
But nothing turned out to be further from the truth. The employees were able to do more interesting work after the digital inspector was introduced. They could think again about solving problems instead of fighting symptoms.
And that is the purpose of the software that BrainCreators makes. Ensure that employees are no longer doing boring, repetitive tasks but can be used for tasks that really need people.
"In the years that we have been working with BrainCreators, I am struck time and time again by how many things people still do by hand," says Wognum. "For example, one time we came to a large carrier and saw that one person could only do the scheduling of all the delivery people for the entire company."
"This person worked daily in a gigantic excel file with all kinds of different tabs to update the planning. And if that person were to drop out or change jobs, the whole company would grind to a halt. And that's just one example. It happens everywhere," says Wognum.
Artificial intelligence is a gray area
BrainCreators' technology actually falls under the heading of artificial intelligence (AI). The company develops its own AI models and turns them into digital inspectors.
For example, BrainCreatos also has AI software, called ARA, that can anonymize individuals on camera footage. There are "digital inspectors" that can recognize what kind of clothing will be popular in the coming season so that fashion companies can adjust their purchasing based on these insights.
"Yet we deliberately choose not to use AI in our communications. Artificial intelligence is too grey an area, meaning customers immediately have very high expectations, while only limited things are possible. Everything depends on what we teach the software to do," Wognum explains.
"So AI, or machine learning, is actually still very limited. We've been waiting for years for a scientific breakthrough that would allow us to use much less data to teach a system what a table is, for example," Wognum says while pointing to a table. "We can recognize what a table is with some explanation, but AI can't do that after one time. It takes hundreds of images to do that."
We can always pull the plug.
Wognum is therefore not at all worried about doomsday images of artificial intelligence starting to think for itself and thereby threatening humanity. "I think the application of AI and machine learning will be used very specifically. A bit in the way we do it, but more and more sophisticated," says Wognum. "Computers are eventually going to take over the boring and repetitive tasks from humans, and I think that's where it will stay. But if it does go wrong, we can always pull the plug."
Wognum is also a connoisseur in this, as he has been working on artificial intelligence for almost thirty years. It started in the 1990s when he was studying artificial intelligence. After completing his studies, he did several projects and developed, among other things, software that can recognize people's outfits in TV series. This information was then linked to products in webshops.
Therefore, it is not surprising that investors knock on BrainCreators' door when they have questions about companies engaged in AI. "These days, we also do due diligence for investors who want to invest in AI-related companies. We then look at the software they are building and assess how well they are doing," says Wognum.
"What we still see time and again is that everything is made on the same principle from 40-50 years ago. Those who claim that their system, with two bicycle images, can recognize all bicycles are actually showing off something that already exists. This software is often already pre-programmed with other examples," explains Wognum.
BrainCreators, meanwhile, is seen as an expert in the field of AI and machine learning. Not because the team advertises it a lot, but because others are knocking on the door of the Amsterdam-based software company. "It's nice to see that we are asked for many things and that our knowledge is recognized, but also that our technology works."
Read the dutch article on business insider