Programme

We’ve clearly seen in the COVID crisis that highly digitized and automatized companies have weathered the crisis better. By intelligently linking and analyzing their data, they could adapt their products and production lines faster to changing needs. Other global challenges like climate change are just as urgent. Here, too, the industry must act quickly by exploiting the infinite opportunities from infinite data. This is where Artificial Intelligence comes into play: it will be an unprecedented amplifier of efficiency and sustainability, due to its capabilities to collect and analyze vast amounts of data and to equip machines and people with digital capabilities.

In this distinguished keynote presentation, Apple board member and SVP Machine Learning and Artificial Intelligence Strategy John Giannandrea will elaborate on the vision and values that drive one of the most important AI players today. 

AI has become a key driver of technology. We are experiencing a “digital” revolution that differs from the first two industrial revolutions in that the processing of information, rather than energy, is being industrialized. Machine learning methods have ushered in a new phase of this development, opening up information sources in unstructured data. The keynote covers the development and successes of AI, and criticizes that current methods are limited by a lack of causal understanding. It also discusses how to keep up and shape this field, including our experience on how to build ecosystems such as Cyber Valley in Germany, and thoughts on how to become internationally competitive by meaningful cooperations among the top European AI sites involving academia, industry, and governments.

Artificial intelligence has become table stakes of modern enterprise applications in recent years. Successful enterprises will become "AI powered" - not just because they can but because they must. Enterprise AI supports companies on their AI adoption journey and harnesses advanced artificial intelligence techniques to drive digital transformation. Join Dr. Feiyu Xu, Global Head of Artificial Intelligence, and Andreas Welsch, Head of Intelligent Processes, SAP S/4HANA, to get firsthand insights into Enterprise AI, and how SAP delivers value by making end-to-end business processes more intelligent.

When talking about data and analytics, we usually think of our main company KPIs. They reflect our success as a business and we want our stakeholders to have easy access and actionable insight from them. But there is so much more to it than just that… Learn how FREE NOW has used the Tableau Platform to drive its business analytics as well as other projects forward even during COVID-19. Business Analytics Lead Michael Thüne shows us how to better understand different aspects of your performance and efficiency - and why taking your data culture to the next level is a significant competitive advantage.

Human-centered robot design, tactile control and physics grounded machine learning algorithms allow robots to acquire new capabilities and skills. This machine intelligence technology will transform society, as we know it today and start becoming relevant to professional service and domestic applications. The recent Corona crisis made clear how much intelligent tactile robotics technology could support us, this applies in particular if it is controllable via AI-enhanced telepresence. Today I would like to give a preview on the potential of this technology and show how intelligent robots will leapfrog us into a future with no unnecessary work commutes, enter environments that are dangerous for humans or provide faster and safer instantaneous medical care.

Powered by AI and the energy from the sun, the Mayflower Autonomous Ship (MAS) will be able to travel further and reveal more about the ocean. Working in tandem with scientists and other autonomous vessels, the crewless Mayflower provides a flexible and cost-effective platform for deepening understanding of critical issues such as global warming, ocean plastic pollution and marine mammal conservation. MAS is led by marine research organisation ProMare with IBM acting as both lead technology partner and lead scientific partner for the project.

The German Federal Government updated its national AI-strategy in December 2020 – two years after its release. What are the main challenges that we are still facing and what is the governmental approach to foster technological innovations while addressing the risks?

Data and Analytics for all!” — The new mantra for today's modern company. However, it is not easy to put all of an organisation’s analytical data and assets into the hands of everyone that needs it! That’s why embarking on this democratization initiative requires you to be prepared to overcome the three monumental challenges you undoubtedly will face. Join Viktor to learn how to:

• Create a semantic layer: Simplify access to the different data assets within the Data Lake for BI-Users & Data Scientists to realise true self-service.

• Ensure Performance: Dremio’s query engine provides a massive boost for BI-Dashboards & ad-hoc analyses

• Avoid Lock-In and Lock Out: Your data in standard open formats, in your accounts - no matter on-premise, cloud, or hybrid - Say no to copies!

The cloud, IoT, machine learning - many new technologies can improve your production and quality assurance. How can you benefit from these new technologies in your business?

Digital Twins are at the heart of the fourth industrial revolution. With rapidly evolving technologies like machine-to-machine communication and the rise of ever more intelligent automation, the basis is set for powerful virtual representations of real-life objects.

The talk presents a vision of digital twins in intelligent automation, which goes far beyond the collection of data on a specific real-life object. Powered by means of artificial intelligence and mathematical modeling, virtual representations are formed that do not only reflect the state of a real-life object but virtually replicate its functionality. This approach enables striking opportunities: Sophisticated analytics, performance optimizations, deep model understanding, and a broad range of options in the modeling of future scenarios are only some of the fascinating possibilities in reach.

The Data Mesh paradigm is a strong candidate to supersede the centralized data lake and data warehouse as the dominant architectural patterns in data and analytics. It promotes the concept of domain-focused Data Products which go beyond sharing of files and towards guarantees of quality and acknowledgment of data ownership.

Through personal experience with applying the Data Mesh concept in practice, as well as dedicated field research, the two presenters discovered the most common pain points at different stages of the journey and identified successful approaches to overcome those challenges. In this talk, you will gain both technical and organizational insights ranging from companies that are just starting to promote a mindset shift of working with data, to companies that are already in the process of transforming their data infrastructure landscape, to advanced companies that are working on federated governance setups for a sustainable data-driven future.

Sometimes it is AI at first glance, but very often it is artificial intelligence only at second glance. Even though artificial intelligence is the spearhead of digitization. Whether in applications of health, the environment or agriculture - it is mainly the things that are completely everyday for us in which AI is hidden. In other areas of life, on the other hand, we do not even suspect it. Antonio Krüger takes a look beyond the obvious.

Big Data brings many benefits to the healthcare industry that can also be used in the context of artificial intelligence and machine learning for the further development of medicine. Nevertheless, the majority of these developments seem to reach for the stars too quickly. The basis, a harmonized network of standardized data in the context of a structured vocabulary, often seems forgotten. Without wanting to slow down medical research and the inclusion of artificial intelligence as well as machine learning, the development of digitized medicine requires small steps in the sense of comprehensive data harmonization. One thing is clear: a long-lasting house can only be built on a stable foundation.

This also applies to the health care system: In the course of digitalization, the amount of data increases more and more. Statistics show that an estimated 2,314 exabytes of data are generated worldwide in the healthcare sector, with data sets growing by 48% per year . In this way, a data pool is created which, however, remains untapped due to the unstructured nature of the data and is therefore not useful for research and development: Institutions that want to electronically evaluate medical data across institutions and sectors require semantic interoperability, otherwise such evaluation will not work. This hurdle, an interoperability of the different systems, can be overcome by responsible data harmonization.

Too often, however, the need for comprehensive terminology has been overlooked in order to semantically encode complex medical issues. After all, the non-systemized nomenclature also contains potential dangers for the patient. The advantages of Big-Data can hardly be mentioned here. The fear of the possible effort is too big. The potential and value of uniform, compatible data are forgotten too quickly.

SNOMED CT, as part of the Unified Medical Language System, offers itself here as a logical terminology standard for coding medical information. With this standard in mind, the disordered data records of various medical institutions can be adapted to this systemized nomenclature. With the help of machine learning we want to facilitate this process: Existing systems of the Meta Data Repositories will be extended by the automatic description of relation based on the SNOMED terminology standard. Using artificial intelligence, the unstructured data sets and information platforms are equated with the SNOMED system. Here, the existing databases are mapped to SNOMED CT using a black box, while at the same time various suggestions for appropriate terminologies are generated for the user in order to take into account the individual and highly specific use cases of the respective institution.

Despite the necessity of this development, the creation of a secure basis for the digital and intelligent progress of medicine – such a system solution does not yet exist. The idea of artificial intelligence and machine learning for the purpose of data harmonization has not yet spread sufficiently. We want to change this as quickly as possible - and we are able to change it!

Applications and Financial Support Plans, short-term Compensations: the Covid crisis has overwhelmed Public Administration in recent months with an unprecedented flood of application processes. In many places, it has been possible to help quickly and non-bureaucratic and to virtualize application processes. But yet Corona points out where the problem is. Especially when it comes to receipts, certificates, documents and correspondence to be submitted, not only the Authorities but also Applicants are often overloaded. The fact that we have merely "electrified" procedures instead of actually digitizing processes is now taking its toll. Andreas Klug heads Bitkom's Artificial Intelligence working group and outlines the prospects of AI for public administration.

In 2017 machine learning algorithms were for the first time able to understand as many words as humans. With this technological breakthrough in 'so called' natural language processing (NLP), digital voice assistants such as Alexa and Google became marketable, with their AI technologies already enabling reliable automation of repetitive tasks in the contact center.

Nicolas Konnerth (Head of Voice @ERGO) and Malte Kosub (Co-founder @ Parloa) talk about this in their keynote "Conversational AI @ ERGO: From an Alexa Speed Boot to a Conversational Strategy". The two share exciting insights with the use of conversational AI in an ERGO contact center; they explain the rapidly increasing relevance of this technology and give an in-person demo of the current technological change for customer service telephony.

By now, almost every company has launched at least one attempt to generate value with the help of AI. Initial proofs of concept often look promising, but unfortunately the big payoff often fails to materialize. Why is that? And how can these pitfalls be overcome?

One key: MLOps. It provides a framework of practices with the whole ML lifecycle and its stakeholders in mind. It sets standards for collaboration, development and management of production ML systems.

In this presentation, Dr. Susan Wegner will give an introduction into MLOps followed by a practical demonstration of the MLops benefits. This will be shown by a real customer project implemented by experienced Data Scientists and Engineers, who already brought several AI applications into productive usage and optimized this technique.

We all search on the web, in intranets, for leisure and work, and on desktops or mobile devices. We all value a smooth search experience. The price we normally pay for this experience is the loss of control over our search-behaviour data. The German privacy tech company Xayn demonstrates with their search engine that AI made in Europe can protect users’ privacy while also guaranteeing a superior search experience. Even better, users can control and interact with that AI which runs entirely on their own devices. This harnesses new business value for AI in search. In this talk, the co-founders Leif-Nissen Lundbæk and Michael Huth discuss the development of the decentralized technology for their search app and present how AI, privacy and user control can be combined. Product video: https://youtu.be/2y6YVGSYEQs

Machine learning can add value in any industry. After all, customers expect information not only in real time, but increasingly also proactively. In addition, one of the biggest challenges for companies is to react to critical moments before it is too late - for example, fraud detection of financial transactions or predictive maintenance in industrial environments.

Apache Kafka is already used as a central nervous system in more and more companies. Machine Learning is an excellent complement! Kafka is used for data integration as well as for training, applying and monitoring analytical models.

This talk will show how you can build a highly scalable and reliable machine learning infrastructure without a data lake. Streaming data is integrated directly via Apache Kafka from various source systems, processed and used for Machine Learning.

The talk compares this modern streaming architecture with traditional batch and Big Data alternatives and explains the advantages and disadvantages of the approaches.

Mindful Chef is a health-focused company that delivers weekly recipe boxes. In order to create a more personalised experience for their customers, they teamed up with Pivigo to develop an innovative recommender system. In this talk, we will tell about this project and the development of a novel approach to understanding user taste that had an unexpectedly large impact on recommendation accuracy. We will further share the challenges of implementing the solution, the impact on the business of the system running live for six months, and further data science projects developed in parallel with implementation.

Financial Planning and Forecasting (FP&F) is key for industrial and financial companies. However, the traditional FP&F approach has several weaknesses, including organizational biases, strive for perfectionism and negotiations between the different entities involved, leading to time-consuming and inefficient processes. Novartis developed an innovative data- and AI-based FP&F approach, replacing the traditional process – a disruptive step towards the vision of the "reimaging business Finance".

While many companies are looking for business cases in finance, the approach Novartis takes represents a successful application of AI in finance. This talk is two-fold: first, we address the technical side and discuss how AI and machine learning can be used for financial planning and forecasting, including inherent challenges. In particular, the tensions between prediction and forecasting, which require causal modelling, are highlighted. Second, we focus on the organizational and practical aspects from data collection, preparation and model creation to the model rollout into individual country entities. We also touch upon the upskilling activities put in place that allow Novartis finance employees to fully leverage the digitalized finance environment. Finally, we will discuss how we see the future of finance in light of AI and the confluence of digital technologies.

For many insurers, the digital transformation is a huge challenge. Customers expect services from their insurance carrier they know from other businesses: all (former manual) services must be available fast, easy and online. To offer their services this way insurers need to transform comprehensively. The necessary evolution of the industry will be fuelled by the integration of automation and AI-powered digital workflows to claims management.

The manual claims processing is a time-consuming and costly procedure, which is being automated with success at Eucon using straightforward process automation, but also machine learning (ML) and other AI methods. We developed a new paradigm of streamlining the claims process, which not only reduces costs but also the claims processing time by an order of magnitude (i.e. from days to less than an hour).

To offer a completely automated process we include AI components in several modules and process steps. Examples are an OCR trained on invoice information, an intelligent document routing system and several prediction services. On the one hand the AI components support our colleagues internally e. g. by minimizing the manual effort of data acquisition. On the other hand, the prediction service is a ML module for our customers giving an estimation on the occurrence of savings potential in the invoices. Using the prediction service, documents with a low probability for savings can be processed very fast without any human interference for the document inspection. This way, our experts can concentrate on documents with a high probability on a likely saving, which optimizes the quality of inspection. In addition, their knowledge is very important for the improvement of the AI modules as AI requires high data quality to learn from and quality checks of already automated processes and decisions.

From a technical point of view, the document classification as well as the prediction service are supervised classification models (e. g. decision forests, logistic regressions or neural networks). Furthermore, unsupervised classifications are used to structure the descriptions of single items on the documents that are used as input for the classification models afterward.

In our talk, we will present how the former very manual process of claims management is automated using ML and AI methods in the above-described modules. Therefore, the internal improvement of case processing as well as the benefits for our customers will be shown. We are currently in contact with some of our customers and expect that one will join us to explain how the prediction service supports them with the digital transformation and how customer satisfaction is increased.

Finally, we will have a short look at how the results of the AI modules can become transparent and understandable using model agnostic algorithms, since our customers and we demand global and individual decisions to be as transparent and explainable as possible. Model agnostic tools as LIME can be applied to regression models, classification models as well as text models and image classifications and provide an overview of the features, that contributed to the decision for individual cases.

AI may be the so far fastest and most broad technology shift, transforming many/the most social, economic and ecological processes. AI can be a part of the solution for some of the most significant global challenges of our time. None of these solutions are feasible without semiconductors, since future AI applications are based on semiconductor solutions. Infineon is already a user and provider of AI solutions. Through system expertise in hardware, software and algorithms, Infineon connects the real world with the digital world and drives innovative solutions that, while taking advantage of these AI technologies, preserve and even extend privacy (e.g. radar detection in secure edge implementation vs. camera based cloud applications), and enhance robustness and energy-efficiency. Edge or embedded AI applications are becoming increasingly important in this context. Edge AI enables data processing with AI close to the sensor without communicating with a cloud and thus makes applications more energy-efficient, faster and safer. This is a special opportunity for European semiconductor manufacturers such as Infineon and the Silicon Saxony industry cluster.

Artificial intelligence will revolutionise the health sector in the next few years. While Vision AI has already received great attention in medicine the great potential of the human voice as a biomarker is still little-known. Dagmar Schuller conducts research in the field of AI, machine learning and big data for more than a decade. With her company audEERING she sets international benchmarks in the field of Audio Intelligence and Emotion AI. Using methods of machine intelligence and deep learning, more than 50 speaker states and emotions as well as other essential characteristics of the human voice can be analysed. The technology thus offers the chance to enormously improve especially the diagnosis and therapy of neurodegenerative diseases such as Parkinson's, Alzheimer's or depression. In her lecture, Dagmar Schuller will explain the basics of AI-based audio analysis and in particular present her latest project: The AISoundLab app. With this app, audEERING has developed an application that uses intelligent speech and sound analysis to indicate abnormalities in breathing sounds and thus contributes to the early detection and risk assessment of COVID-19.

Every year, hundreds of new algorithms are published in the field of biomedical image analysis. For a long time, validation and evaluation of new methods was based on the authors’ personal data sets, rendering fair and direct comparison of solutions impossible. In the meantime, common research practice has changed and involves organization of international challenges that allow for benchmarking algorithms on publicly released data sets. After more than 15 years of biomedical challenges, we have carefully analyzed and critically questioned common practice related to the design and organization of biomedical image analysis challenges. The talk will provide statistics and new insights related to various topics around biomedical image analysis challenges, ranging from the choice of metrics and ranking schemes to the pitfalls that may occur when organizing or participating in a challenge.

The presentation shows how the investment in advanced analytics helps to decrease the operational costs by reducing the manual monitoring work and to minimize plant downtime by predictive maintenance. Efforts to enhance analytics in the utility sector often rely solely on applying machine learning techniques to data readily available from the machine control (DCS). It is often observed that the complexity of the involved machines and processes cannot be fully covered by using readily available machine learning algorithms but needs to be tackled by advanced hybrid physics based models.

In the field of artificial intelligence (AI), machine learning (ML) methods have recently led to enormous advances in automatic pattern recognition and processing of complex data. The application of machine learning methods is particularly useful when cause-effect relationships are difficult or impossible to describe analytically with mathematical models, but instead extensive data is available. This situation is often encountered in manufacturing and industrial prodution. Production plants are extensively monitored by means of many sensor devices so that ML processes can be initiated, action plans can be generated and subsequently executed, resulting in continuous optimizations of production processes. The AI Innovation Center »Learning Systems« at Fraunhofer IPA is part of the Cyber Valley, which is Europes largest research cooperation in AI. The AI Innocation Center supports companies and especially small and medium-sized enterprises in the field of AI and ML for manufacturing with a special focus on the transfer from basic research to industrial application. It offers various forms of cooperation: »Quick Checks« are analysing the feasibility of solving a pratical manufacturing problem by means of AI, while »Exploring Projects« turn the findings of a Quick Check into a prototype.

This presentation will provide insights into technology transfer done at the AI Innovation Center and experiences from the various collaborations with manufacturing companies. The use cases presented range from quality control to robotics to explainable AI. It is intended to demonstrate the benefits of AI in industrial production and to reduce shyness when tackling and implementing AI applications.

As in other areas of business, artificial intelligence becomes increasingly important in the banking sector. As of today, deep learning and machine learning are the core pillars of artificial intelligence. Thus, the banking industry needs to use the significant developments in these technologies which can help improve the understanding of banking processes, types of transactions as well as data-based decisions. As use cases based on machine learning are further developed, tested and implemented, the magnitude of this technology becomes obvious and more practical. In his keynote Dominik Schmidt-Kiefer will introduce you to the possibilities artificial intelligence has for the banking sector. Further, he will demonstrate the practical implementation of machine learning in the banking sector based on an actual use case where the loan decision is supported by a neural net, automatically tagging account transactions. He will give insights in the development and implementation of the use case and share learnings.

For Lower Saxony their cadastral maps (Katasterkarten) are essential to support planning, taxing, construction and administration. The current maps however are lacking consistency. Some areas have been mapped by hand in the late 1800s and then copied and digitized. Some areas have been mapped today with high precision GPS. In order to homogenize the maps, LGLN is using artificial intelligence to identify objects in orthophotos, to detect the outline of buildings and to do a comparison with the cadastral maps. We give some insights into the project setup, the technology and the successes of this project.

In this talk, Melinda will explore the ethics of immersive design and "gamification”. Games are powerful experiences that immerse people in new environments and engage people in the stories that happen within these universes. More and more we look to games for best practices and philosophies to try and apply these design methodologies into more business-minded - not entertainment alone - contexts. This, of course, opens the door for big questions concerning the ethics of privacy, surveillance, and monitoring when organizations deploy digital game-inspired tools to collect and use data for customer, employee, and stakeholder engagement.

Melinda will discuss topics such as how the bandwidth for immersion has changed with new technologies, the danger of designing for “untethered metrics”, uncovering whose goals the UX is honestly intended for, why "fun” isn't a value proposition on its own, that entertainment isn't a fair trade for manipulation, and where the responsibility lies in addressing ethical implementations. She'll share why she believes ethical design is about understanding your role as a guide, not an enforcer, and accepting the responsibility that comes with using technology in a way that allows you to reach, connect, and engage with your audience in more intimate and leading ways than ever before.