Come again? Test your digital vocabulary. No. 2

The idea of ‘deep learning’ has been around since the 1950s. To enable machines to learn, artificial neural networks are programmed, which simulate functions of the brain – or more precisely the cerebral cortex. This is the part of the brain where thoughts, associations, comprehension and sensory and motor activity are controlled.

As a special area of machine learning, deep learning is now one of the most important research fields on artificial intelligence. Google’s Android Assistant, IBM’s Watson and Microsoft’s Bing Voice Search all use deep learning for better speech and image recognition and more meaningful forecasts and decisions. Google works and researches with one of the biggest deep learning systems in existence. In 2013, the company presented a neural network made up of more than one billion connections and 16,000 computer processors. The system was shown ten million images from randomly selected YouTube videos, which it was able to sort into 22,000 categories. When it came to identifying objects, such as cats or children, the result was twice as good as that achieved by previous systems.

In China, Microsoft demonstrated a voice recognition tool based on deep learning, which transcribed the spoken word with an error rate of just seven per cent. The voice recognition tool then translated the text into Chinese and simulated the original voice of the scientists when reading out the text in Mandarin. That is impressive, but it is of course still a long way from the complexity of human thought.

Even more new concepts, software and processor performance are needed to get close to this, which could take years, maybe even decades. Yet the progress that is now possible with deep learning will change many things. The concept is suitable for all tasks which involve large amounts of data in which patterns and models can be found. IBM’s Watson, for instance, helps doctors with diagnostics and can already give a more precise diagnosis than some doctors. Autonomous driving also uses deep learning for ever better image recognition with the aim of improving safety.

This is another idea that has been around for a few years and is now taking off. It is only worth creating a digital twin with the Internet of Things (IoT). Some people say that companies have to work with digital twins nowadays.

A digital twin is a virtual model of a process, product or service that really exists. As a result, the model can be fed with real data – such as running time, position, precision, speed, temperature and much more besides – throughout the entire life cycle, from the design and manufacture to the operation. This data is recorded on the actual product or process using smart sensors and sent in real time via the cloud. In conjunction with other measured values and factors, it is thus possible to use the model to identify, simulate and test risks, weaknesses, changes and opportunities, allowing systems to be monitored and vulnerabilities even eliminated before they occur. This has many advantages for the new connected products and services in particular.

For a long time, NASA has been using the concept to completely design, test and build virtual rockets to begin with. Only once the digital twin has perfectly satisfied all requirements will the real rocket be built. Using smart sensors, all components then once again become data sources for the digital twin in order to test how they work in reality. The Port of Rotterdam, Europe’s largest container handling facility, has commissioned a digital twin of its infrastructure. Around 140,000 ships are loaded and unloaded there annually. Rotterdam now wants to become the ‘smartest port in the world’. An IoT network with sensors collects real-time information on all possible conditions across the entire site. This is correlated on the dashboard with other data, such as the weather and water level. The idea is make the port management operation safer and more effective.

Behind the acronym SMACT are five technologies that are fundamental to companies in the digital transformation process. They are Social Media, Mobile, Analytics/Big Data, Cloud Computing and the Internet of Things.

They may not all be brand new, but when combined they take on a new importance. The fusion of the initials of the various technologies to form one word is designed to make this clear. Together, in the way they interact with one another, the technologies can be used effectively and progressively within companies. Here, too, the Internet of Things (IoT) is an important driver of development.

Also new is the fact that these technologies are not necessarily based in the IT department, but are managed by the marketing team. That’s because all five technologies help to improve the customer relationship in various way: as a way of engaging in dialogue, as a service and convenience, as a research lab to analyse customer behaviour and as a network to encourage better products and services. It goes without saying that reaction speed and the evaluation of results play a major role in the use of these apps or software, but anyone who masters these aspects has already successfully implemented the digitalisation process and will find new ideas for the future of their company.

Here is a fun fact for an intellectually stimulating conversation: it is possible to study for a degree in SMACT. Then graduates would be a Master of Science in Art, Culture and Technology.