Artificial intelligence

When machines think

01/2017 - Whether Siri from Apple or Echo from Amazon, smart assistance systems are finding their way into our everyday lives. At DB Systel too, the topic of artificial intelligence is given high priority. In this interview, Rüdiger Kurz and Jens Pfeiffer talk about the initial successes and the next stages.

digital spirit: What actually is artificial intelligence?

Rüdiger Kurz: Fundamentally, artificial intelligence – or AI – is concerned with the automation of intelligent human behaviour. Apart from the recognition of images and objects, the planning of actions and deduction, this also includes language processing and mechanisms for machine learning. Specifically in these areas – natural language processing and machine learning – AI supports knowledge management and thus contributes toward speeding up the process. For example, AI can reduce the need for cognitive effort when scanning a text by automatically creating a summary of the document content.

digital spirit: This definition is not particularly new, is it?

Jens Pfeiffer: I learned the basics of this during my studies back in the early 1990s. The changes since then have been barely perceptible. But now AI is actually achieving added value.

digital spirit: Why only now?

Jens Pfeiffer: New algorithms, storage of graphs, improved database query languages and a high degree of standardisation, combined with new computing technology mean that we can now talk of a technological leap forward in the field of AI. Many people and companies are currently occupied with this exciting topic.

Rüdiger Kurz: Artificial intelligence is a key topic everywhere at the moment and is being promoted to a great extent by the Big Five, that is Microsoft, Google, IBM, Apple and Amazon. The continuing technological developments have triggered a certain coverage of the topic at Deutsche Bahn as well.

digital spirit: To what extent?

Jens Pfeiffer: The amount of development is enormous. In the USA, AI services are already being used in hospitals. A Watson system is performing an advisory role in diagnoses: it is fed with millions of data records about disease patterns and can analyse these better and faster than a doctor could. Artificial intelligence can therefore deal with data very effectively – and not just in medicine, of course. You can easily imagine how the preparation of a railway timetable can be made more efficient with AI.

© DB Systel GmbH

Many processes at Deutsche Bahn, for example in the processing of queries in the dialogue with customers, in the management of contracts or safety, can be supported or even automated using AI technologies. Customers and employees will benefit from a new range of added value features.

Rüdiger Kurz, expert for Open Source and AI at DB Systel

digital spirit: Is there a special AI unit at Deutsche Bahn that is dealing with this topic?

Jens Pfeiffer: At present, work is under way on a universal AI strategy and it remains undecided whether there will be a dedicated AI team. The generic AI technology can, however, be used in a wide range of scenarios. But DB Systel has already established an AI community that exchanges information on a regular basis. All members of the group are warmly invited to participate as active or interested community members.

digital spirit: What is the objective of using AI?

© DB Systel GmbH

It is a matter of making the input as easy as possible and the output of successful hits as precise as possible. This is what AI is there for: to use assistance systems for greater convenience so that not everything has to be input individually, but instead the machine can autonomously gather the necessary information and the context – and thus achieve a good quota of hits.

Jens Pfeiffer, AI technology mentor and trainer of Pepper at DB Systel.

digital spirit: What applications has AI already been used in?

Rüdiger Kurz: Within nine months we have implemented three pilot projects – from securing orders to their implementation. For example in the case of the “Reisebuddy” (Travel Buddy) application and for the “Laura” assistance system. AI has also been deployed in the “Pepper” robot. And for DB JobService we have developed the “Kompetenzmarktplatz+” (Competence Market Place) in order to identify job prospects for every employee throughout the Group.

digital spirit: What exactly is the role of AI here?

Rüdiger Kurz: On the basis of employee competencies, similar and related activities in the entire Group are ascertained in order to speed up a placement in a new activity. In the Kompetenzmarktplatz+ existing CVs are read automatically as text or graphic documents. A linked search for open positions, as well as tasks and qualifications makes the job offers within the Group more visible and reduces barriers to the job market within the Group.

digital spirit: How does AI work in practice?

Jens Pfeiffer: Take the example of the IT Helpdesk. This is a single point of contact for all Group employees, where they can discuss problems in their everyday work on a hotline. Examples of typical problems reported are: my computer will not boot up, I can’t send any e-mails or I can’t connect my mobile phone to the WLAN. And a wide variety of similar problems. The Laura system recognises these questions semantically and can answer them in fully or semi-automated mode.

digital spirit: What can you teach an AI system?

Rüdiger Kurz: Based on training data, models are set up with which patterns are recognised, classifications carried out or decisions made based on values gained from experience. Let us take the example of the “Reisebuddy” travel assistant. Say a customer wishes like to travel by train to Berlin. He can formulate this request in many different ways, whereby the language, the completeness and the differentiation from other intentions play an important role. In order now to enable the machine to answer a colloquially formulated inquiry, it must be trained using previously asked questions and appropriate answers. By means of pattern recognition, the syntax and formulations which were not part of the training data can be correctly interpreted. The higher the quality of the training data, the better the model and the smarter the AI. In the case of Reisebuddy, the machine can, for example, ask for missing information such as the departure location and then direct the background systems using the completed and structured query in order to offer the passenger the best option.

digital spirit: Can these products also be deployed in other areas?

Rüdiger Kurz: Work is already under way on this: DB Station&Service is interested in a very similar use case. Overall, we want to promote the reusability of individual components, so in the long term we can establish a generic solution that we can adapt for different customer issues.

digital spirit: What effects does it have?

Rüdiger Kurz: AI is indirectly an accelerator, whose use can cut costs such as external personnel costs. By being able to speed up processes using AI, however, first and foremost we improve the satisfaction level, the user experience and the quality of service for rail customers.

digital spirit: Many thanks for talking to us.