Binary District Journal spoke with Matthias Gemmar, Head of Strategy at ITK Engineering, about Bosch’s plans for IoT, his involvement in their operations, and how machine learning is integral to the future success of the technology.
Industrial IoT adoption is increasing rapidly. In the IIoT Maturity Study, which tracks the adoption of IoT in the manufacturing, transportation, and oil and gas industries, 86% of industrial organisations said they are in the process of adopting IoT solutions. As many as 84% of those believe that these solutions are ‘very or extremely effective’.
“86% of industrial organisations are in the process of adopting IoT solutions, and as many as 84% of those believe that these solutions are ‘very or extremely effective’.”
The benefits of the technology are many. The insights yielded through analysis of the data collected by connected devices have the potential to be transformative for business. However, this data will arrive at an unprecedented rate - even for companies who would consider themselves in the upper echelons of data maturity.
It's analysis cannot be done by humans. The process requires emerging technologies such as machine learning, a technology that is fast approaching mass adoption. And there are other issues.
Power to the People
One of the reservations holding the development of IoT technology back is a lack of immediately obvious real-world business use-cases. Outside of manufacturing, which already has clear use for sensors and data collection, and software companies, these use-cases will take time to become apparent.
Where these solutions do emerge, you then need companies able to bring them to fruition. ITK Engineering is one of those companies, building solutions across a variety of use-cases. According to Matthias, ITK works to provide ‘white box’ solutions, where the intellectual property of the software belongs to their customers.
“Customers bring their problems to us and we work on their problems, not our own,” he explains. “We are very strong when it comes to software and system development for very complex systems.
“We have developed an algorithm that transfers from the sensors on this system to the machinery. This makes the doctors feel as if they’re actually touching the human body looking for tumours.”
“For instance, we have a system where we show haptic feedback – something very complex on control systems. We have developed an algorithm that transfers from the sensors on this system to the machinery. This makes the doctors feel as if they’re actually touching the human body looking for tumours. This is a complex task for the algorithm because we have to transform what a sensor ‘feels’ into the movement of the machine.”
In essence, they are using advanced control systems as well as advanced requirement engineering to solve challenging problems for their customers. The ability to help doctors locate and identify tumours might not be an immediately obvious application of IoT technology, but it is an example of the myriad solutions that make the potential of the technology so exciting.
The Algorithm, the Algorithm, the Algorithm…
For ITK, the focus is all on the algorithm - they leave it to their partners to create the device.
On how it all comes to fruition, Matthias says that they have a system engineering approach. ITK locks in the requirements with its customer and puts them into a system design so the customer knows how it is linked. It is then that they look for partners who can give them the sensors they need.
Thanks to Bosch, ITK has a host of sensors available which they can use and link into the system. As Matthias puts it, “My main focus is the algorithm that makes the thing intelligent.”
Ultimately, without machine learning the IoT is useless in a business sense. Yes, connected devices can be used by companies to collect an incredible amount of data. A series of sensors along the inside of some industrial piping, for example, will be able to pick up hitherto inaccessible data, data that is potentially extremely valuable to those using it.
Without machine learning, though, this data will be all but unreadable. It will be messy, there will be far too much of it for humans to derive any meaningful value from it, and generating actionable insight will be difficult to the point of not being worthwhile.
Living Together In Harmony
Since this problem-solving requires close interaction with their customer base, ITK begins working with them right from the product’s inception. “They don’t come to us and say ‘go program it’,” Matthias says. “We try to develop the idea to a product from the very early stages, so that we have teamwork with our customers.”
ITK coordinates by working openly with its customers and taking regular feedback. The process also involves regular prototype building that helps both ITK and its customers see that they are going in the right direction.
This kind of coordination will be key in developing meaningful, functional IoT solutions. As is the case with a lot of technological developments, those developing the tech have no experience of the field they’re developing it for. Therefore, a continual back and forth will be required between programmers and those in the field to build workable products, at least initially.
The Hunt for Use Cases
ITK’s approach to IoT and machine learning involves working on embedded systems and big machines, which it has been doing for a long time.
“With IoT, it’s about connecting machines, and there are not enough standards or common bases on how to connect things,” Matthias says. “It’s a lot about use-case thinking; it’s about building up use-cases with customers. Right now, we’re focused on helping our customers to have the right architecture to be able to connect to others.”
“Our focus in consulting is that we only consult on things that we are able to develop.”
This is a step towards IoT readiness. ITK is preparing clients to take the next step to launch their own vision. One of the solutions that the company provides is a self-analysing algorithm.
ITK has a strong base in the development of algorithms, but is realistic about what it can and cannot produce to a workable standard. As Matthias explains: “Our focus in consulting is that we only consult on things that we are able to develop, so we aren't consulting on things that we don't know how to develop.”
The Next Big Thing Won’t be One Big Thing
Matthias does not think the next big development in machine learning will be a single, defined, ‘big thing’. “There will be a lot of things happening at the same time that probably – and this is only my personal view – will lead to the creation of spaces that are very bespoke, that are very customer-specific,” he says.
“Things will be developed into standardisation and a lot of industries will actually come together – robotics is coming in, algorithms, AI.”
Indeed, machine learning and IoT will both have ‘made it’ when they are being used across industries and when those using them barely even recognise that they are. Seamless integration of machine learning into business practices is the goal, not wholesale change and the overhauling of every industry under the sun.
Not every application of the combined power of IoT and machine learning will be inherently positive. When the military’s desire for automation is taken into account, the scope for some truly terrifying machines to be built becomes apparent.
Take Russian weapons manufacturer Kalashnikov. It announced last year that it was working on a “combat module”, an automated machine gun capable of firing on humans through the use of an onboard camera, a computer, and artificial intelligence. Apparently it will be able to ‘learn from its mistakes’ to be become better and better. The real question is, should it be given the opportunity to make mistakes at all? Have they not seen The Terminator?
This article was prepared with the help of the SLUSH conference team and co-authored with BDJ tech writer Shivdeep Dhaliwal.
Illustrations by Kseniya Forbender
To contact the editor responsible for this story:
Margarita Khartanovich at [email protected]
- Artificial Intelligence Isn’t Ready to Take Over From Doctors and Nurses, Just Yet
- Machine Learning Vs. Analysts: Will AI Eventually Replace Data Scientists?
- Why The Danger of Deepfakes Is No Danger At All
- Sarcastic Robots? How Deep Convolutional Neural Networks Are Making AI Worryingly Human
- Traffic Lights in The Sky: Flying Cars Can Appear Sooner Than You Think