<img height="1" width="1" style="display:none;" alt="" src="https://dc.ads.linkedin.com/collect/?pid=1007900&amp;fmt=gif">
en | de

LEAD Innovation Blog

Read our latest articles on innovation management and innovation in a wide range of industries.

Date: 18-Jan-2019
Posted by: Franz Emprechtinger

Why Machine Learning will change the industry from scratch


Many things can be programmed, but not everything. Because some things are too complicated to fit into fixed rules. Thanks to machine learning, computers can also handle complex processes and even unforeseen events. Read in this blog post how this technology can increase the competitiveness of the industry, what role Google, Amazon, Microsoft, IBM and the like play in this, and how machine learning can suddenly turn industrial companies into IT service providers.

Computers need programs. Programs in turn need fixed rules: if case A occurs, then start process B. Computers understand only clear, unambiguous instructions a la binary code. Either 0 or 1 - there is nothing in between.


There are no rules for everything

Complex tasks, such as the recognition of images or the understanding and translation of language, can hardly be pressed into such fixed rules. Programming is therefore not sufficient for these applications. Nevertheless, computers can acquire these abilities. By learning just as we humans do.

We all don't know what a cucumber is from our first cry. We had to learn that first. By looking at or even touching hundreds or even thousands of cucumbers. And our parents told us, "This is a cucumber. Until we understand that.

Trend Collection Innovation management


Computers learn like people learn

A computer learns in the same way. Feed him thousands of pictures of cucumbers and tell him it's one. The more cucumber pictures the computer has "seen", the higher the probability that when looking at a cucumber picture he himself knows that a cucumber can be seen on it. You will be surprised what you can do with a computer that has acquired this knowledge:


Japanese realized self-learning cucumber sorting machine

A Japanese engineer has trained a computer so profoundly that it is now able to sort the freshly harvested vegetables into 9 different quality levels. So far this somewhat monotonous but important work had done the mother of the engineer. Mrs. Makoto ran a cucumber farm with her husband. The son, Koike Makoto, developed the self-learning sorting machine during a longer stay on his parents' farm.


The sorting machine is interesting not only because of its intelligence. But also because of their design concept. The heart of the machine, the computing power required for machine learning, is not located on the cucumber farm. Koike Makoto gets them from the cloud. In the specific case from the one by Google. The server landscape necessary for training the system would have become too expensive. Koike Makoto gets the computing power from Google's data centers, but only pays for the resources he really needs.


Machine Learning from the socket

Similar services like Google's also offer other IT giants such as Amazon, or Microsoft or IBM. The names for this are Google Cloud Platform, Amazon Web Services, Microsoft Azure and IBM Blue Mix. They all offer Machine Learning as a Service. This means that you can implement machine learning in your factory relatively easily: You simply get the necessary computing power from the cloud of one of these providers - in other words, from the power socket. Ready-to-use application modules are also available. For example, a machine learning application that converts spoken language into text and also translates it into 80 different languages.


Industrial companies become IT service providers

You can also provide the IT resources required for machine learning yourself. Either you expand your own IT infrastructure accordingly. Or you choose a hybrid approach. Bosch, for example, has decided to invest in its own IT resources. These are so powerful that Bosch has been making them available to other companies since 2017. Bosch has thus advanced to become an IT service provider. This is a path that other companies have already taken. Amazon, for example: the company, which started out as an online bookseller, needed a fail-safe IT infrastructure to ensure that the online shop always worked. Amazon set up a globally distributed data center and developed highly available services for this purpose. In 2006, Amazon founded its own subsidiary to market these services. Amazon Web Services is now the largest cloud provider and also counts large companies such as Dropbox and Netflix among its customers.


Machine Learning makes production more flexible, faster and more efficient

For the industry as a user of machine learning, the technology offers numerous advantages. Intelligent machines can make production more efficient, flexible, faster and thus more competitive. The following two application examples will give you a first impression of the potential of this technology:


1) Predictive maintenance keeps your machinery in perfect working order

Real and virtual sensors collect numerous data about your machinery during production. Based on data from the past and real-time data, Machine Learning allows you to predict exactly when certain components will wear out. Or how long a component that no longer works one hundred percent will last. With this knowledge, you can service your machine when it is most convenient for you. For example, if the entire plant is standing during a public holiday. However, you can also dissolve your spare parts stocks. Because you know in advance when which component will be broken. You can therefore order it again so that it arrives at you at exactly the right time. Moreover, well-maintained systems have a much longer service life than those that are standing up again and again due to defective parts. With Predictive Maintenance, you can turn many screws to improve your systems.


2) Effective protection against IT threats

Everything that is networked can also be hacked. In the age of Industry 4.0, however, this is exactly what is being worked on feverishly. Security problems are particularly unpleasant in industry. Because a production line downtime of just a few hours means damage of hundreds of thousands or even millions of euros. The current hacker attack shows very clearly how vulnerable the networked world is. However, complex systems can hardly be sufficiently secured with traditional security concepts. Machine Learning can also help here. Simplified this works as follows: Machine learning algorithms search for known patterns that indicate malware. Or you can detect anomalies in the system and sound the alarm in time.

Read our article on Why Industry 4.0 needs robust networks and new mobile communications standards.  


Conclusion: How machine learning is changing the industry from scratch

Thanks to machine learning, the computer could lose the nimbus of the "tinfoot", who doesn't understand anything but literally always proceeds logically according to program. Because technology makes computers capable of learning. Thanks to this ability, the machines will not now take over the world. But this enables computers to master tasks that actually require a little bit of brains. Which, however, basically make people seem monotonous. Cucumbers sort, for example.

Machine learning will change the industry in a very short time. Not only because it makes production faster and more efficient. But also because the roles of the actors are changing very much: Amazon or Bosch are just two examples. Because they themselves have developed something that they urgently need for their daily business, they have become specialists in completely different fields. Such metamorphoses are also possible in completely different directions through machine learning. It would be conceivable if a production company were to concentrate fully on the Smarte Factory and then make it available to other companies that do not own one. Production as a service, so to speak.

Trend Collection Innovation Management

Franz Emprechtinger

Born in Ried im Innkreis. As former Head of Innovation, he was responsible for the entire project management and specializes in the areas of fuzzy front end and business model innovation.

You want to work with us?

We would be pleased to advise you on a possible cooperation to make your innovation management future-proof.

Contact us