By Damir Dobric
Machine learning (ML) and artificial intelligence (AI) are often used as synonyms and are therefore generally confused.
AI was established in 1956 and describes an attempt at emulating human intelligence in an artificial manner. ML, on the other hand, is the first step towards AI. The current hype around AI is hiding the truth of the matter, namely that we are still a great way off a genuine AI.
However, there is one piece of good news: Machine learning already works! ML algorithms require data for this, as they use it to discover patterns. The real art is turning this into a business idea, or in other words, having an idea, having the necessary data and an expert who can evaluate the data.
If a company does not possess the specialist knowledge required for this, it is recommended and common practice to purchase the expertise from an external source. This requires the expertise required by the company to be clarified, as well as the necessary duration of the services.
The answer depends on many different factors: If a suitable ML subscription service is available on the market, the company does not require any additional expertise. However, a subscription service is, in the broadest sense, a cloud model, and therefore leaves something of an aftertaste with many mechanical engineering companies. Would the company become reliant on this service?
This is a question each company must answer for itself. Taking a look outside of the box may be helpful, however. For example, how dependent is Netflix on Amazon Web Services, or Uber on Microsoft Azure? Many may view this dependence as almost negligent, yet the business results of both companies speak for themselves. In addition, both are viewed as "cornerstones" of the digital transformation, with mechanical engineering companies also following in their footsteps despite their different business models.
In the age of the cloud, the integration of various external services is viewed as being not only efficient, but also a way to promote innovation. "Platform as a Service", or PaaS for short, is the name of this approach. And it has a real advantage: Mechanical engineering companies can procure different services from different providers.
Can cloud technology always help? Different cloud providers have been attempting to offer reusable ML services as PaaS for many years. However, their focus is on offers which can be reduced to the lowest common denominator. Customers should therefore not expect these offers to include predictive maintenance for their machines, for example.
The majority of ML services can be placed into the categories of image, video and language analysis, knowledge and search. Mechanical engineering companies can also be a provider of a service like this in their (new) business area.
And how does this work without a cloud?
Cloud technology enables the provision of low-cost, scalable and simple solutions - provided the company has permanent access to the internet. If this is not an option, on-premise cloud services can often be used in theory. Should the company wish to remain independent from the provider of the cloud, there is the option of commissioning a different company to develop this software (this ML service) exclusively for the company. The client (mechanical engineering company) retains the rights to this software and the commissioned company will continue developing the software and making adjustments if desired. These generally small software companies or start-ups are currently in high demand in the market and are often bought for good reasons.
If none of these options is tenable, companies must begin building the necessary skill sets themselves with the help of data scientists and cloud solution architects. The former understand how data models are created, while the latter implement cutting-edge applications derived from these models and integrate the applications in the existing software environment.
Soon, technologies such as cloud, IoT, big data, and presumably blockchain and AI, will decisively change the mindset of companies. Business models will be developed in a more interdisciplinary manner, companies will use external services more often and also provide their own services. PaaS in particular makes it possible to implement these business models in a simple way.
Products will become ever more short-lived and skills will become more complex in the business world of tomorrow. Therefore, it cannot be expected that mechanical engineering companies will become AI experts in general; rather, companies must learn to use external data more often, as well as providing their own data to others. For this to become the norm, we must reduce all our “data fears” to a reasonable level, as only then will the mechanical engineering industry be able to successfully use ML technology.