Loading ePaper
Deutsches Forschungszentrum für Künstliche Intelligenz German Research Center for Artificial Intelligence next AI as a Transformer How Artificial Intelligence is Connecting Science, Business, and Society in New Ways
About us Human-Centric AI Intelligent Solutions for the Knowledge Society DFKI has more than 35 years of experience in human-centric AI, conducting research in key areas of basic research and future- oriented applications, while always maintaining a focus on social relevance and scientific excellence in the field of artificial intelligence. The German Research Center for Artificial Intelligence (DFKI) was founded in 1988 as a non-profit public-private partnership. It operates facilities in Kaiserslautern, Saarbrücken, Bremen, Lower Saxony (Osnabrück and Oldenburg), Darmstadt, with laboratories in Berlin, Lübeck, and a branch office in Trier. It supports 29 research areas, ten competence centers, and eight living labs. Based on application-oriented research, it develops product functions, prototypes, and patentable solutions in the information and communication technology (ICT) sector. Funding is provided by public funding sources and industrial development contracts. Project results and milestones are periodically reviewed institutionally and by a Scientific Advisory Board, an international panel of experts. In addition to the German Federal Ministry of Research, Technology and Space (BMFTR) and the federal states of Rhineland-Palatinate, Saarland, Bremen, and Lower Saxony, several leading German and international high-tech companies from a wide range of industries also have seats on the board. www.dfki.de
Editorial AI as a transformer – Outstanding research for tomorrow’s economy The German government’s High-Tech Agenda identifies AI as a vital technology for the future, and the aim of its AI strategy is for AI to be responsible for ten percent of Germany’s economic output by 2030. Achieving this goal will require bold investment in computing infrastructure and a strong research focus on knowledge transfer. When research results appear not only in trade journals but also in production halls, logistics centers, and medium-sized companies, theory has become innovation, and AI has become a cornerstone of growth and profit. On the basis of our application-oriented basic research, our results are already up and running to develop prod- uct functions, prototypes, and patentable solutions. The feature topics in this issue represent our diverse transfer projects, i.e., Yield Consortium, Carve-DL, ESCADE, PAIRS, Open6GHub+ as well as our activities in the field of biomedicine. Prof. Dr. Antonio Krüger Germany is unique in combining scientific expertise, a strong indus- CEO trial base, and a broad range of small and medium-sized enterprises – ideal conditions for close linkage between science and industry. DFKI promotes transfer through joint research projects, open in- novation platforms, and TransferLabs. In this issue’s interview, my colleague Prof. Andreas Dengel and I discuss why DFKI is a valuable partner for companies in the AI transformation and what the collaborative possibilities exist to work with DFKI. With our diverse industrial partnerships, we create a cycle in which science provides the impetus, industry scales it up, and both sides learn from each other. In this way, AI is not an end in itself, but a means to an end, an AI “transformer” helping to make Germany fit for the future. 3
AI Innovation Yield Consortium: Accurate harvest forecasts with satellite data and AI Agricultural yields can be significantly impacted by environmental factors and un- controllable events. It is rarely possible to predict harvests with absolute certainty. However, this uncertainty is now being turned into assured certainty by the Yield Con- sortium at DFKI. An AI-supported technology platform there analyzes multispectral satellite data to generate precise and reliable harvest forecasts – months in advance. i b a l a d e m G T A / A S E © 4
AI Innovation From satellite imagery to forecasts Farmers often face the challenge of having to make de- cisions under conditions of great uncertainty. The soil is prepared, the seeds are sown – and then, all that remains is to wait. AI is fundamentally changing this pattern: the Yield Consortium uses systematic imag- es from Sentinel-2 satellites from the European Space Agency’s Copernicus program, whose high-resolution multispectral cameras capture images of the Earth’s land masses every five days. “ The Yield Consortium demonstrates how com- plex satellite data and AI-based analytics are used to generate reliable harvest fore- casts, giving farmers and the entire value chain far more planning certainty and stabi- lity.“ Prof. Andreas Dengel Data diversity for better accuracy Satellite imagery forms the core of these analyses. They are supplemented by weather data, soil quality maps, topographical information, and historical crop data. AI is able to identify complex relationships between envi- ronmental factors and plant growth. Test results show a forecast accuracy of over 90% for wheat crops in the US – almost perfect agreement between forecasted and actual yield. 5 OpticalWeatherTopographySoil
AI Innovation “ Yield modeling in conjunction with BBCH and LAI modeling is a complementary compo- Research meets industry The Yield Consortium brings together science with in- nent of crop modeling. BASF Digital Farming dustry partners such as BASF Digital Farming, John GmbH has a strong interest in the diverse Deere, and Munich Re. Close collaboration ensures that yield modeling approaches developed by the findings do not remain in the laboratory, but flow directly Yield Consortium.“ BASF into practical applications. AI reveals the invisible A specially developed model classifies different crops “ The combination of expert knowledge from various industries, from agriculture to in- and forecasts the expected yields. Comparative data surance, with DFKI‘s outstanding data sci- from agricultural machinery confirmed the results. The ence expertise proved to be very construc- field, but also the entire region is made more trans- tive. Together, we were able to improve parent: is there a bottleneck somewhere, or where can overproduction be expected? the predictive power of previous models.“ Jürgen Wendlandt, Head of Data at Munich Re Added value for agriculture and society Accurate forecasting not only helps farmers plan their Knowledge clearly communicated The accompanying video illustrates operations more efficiently, apply targeted use of re- the path of satellite data flowing into sources, and minimize risk. The impact is more far- AI analysis to provide on-site fore- reaching as aggregated data supports food security, casts. It shows how abstract data is trade strategies, and climate adaptation — locally, re- used to create tangible added value gionally, and globally. for agriculture and industry. 6
AI Innovation Outlook: data-driven farming The interaction of AI and Earth observation systems flexible transfer to other cultures and regions and paves creates added value in the agricultural sector. Our AI the way for application development beyond agriculture, technology platform has a modular design that enables for example, in climate observation and urban planning. Prof. Andreas Dengel is Executive Director at the German Research Center for Artificial In- telligence (DFKI) in Kaiserslautern and Head of the Smart Data & Knowledge Services re- search department. His research in artificial intelligence focuses on a wide range of top- ics, including machine learning, image pro- cessing, knowledge services, and semantic technologies. Andreas Dengel’s experience allows him to serve as an intermediary between research and application, facilitating the transfer of innovative AI technologies to various areas — from manufacturing to climate and terrestri- al monitoring, to medical and societal chal- lenges. Teams at DFKI under his leadership develop smart solutions able to analyze complex data sets to discover new insights. This interdisciplinary approach makes him a key spokesperson for the responsible use of AI in science and industry. Prof. Dr. Prof. h.c. Andreas Dengel Executive Director DFKI Kaiserslautern 7
AI Innovation AI in Biomedicine: Insight into the building blocks of life Artificial intelligence is fundamentally changing biomedical research: it is bringing previously hidden insights about the human body to light. Thanks to AI, medicine is becoming more individualized, effective and cost-efficient, with new therapies, cus- tomized medications and preventive measures. At DFKI in Kaiserslautern, our team demonstrates how interdisciplinary research and practical knowledge transfer work together to gain innovative medical insights and save lives. Understand complex data and create knowledge The heterogeneity and complexity of biomedical data “ Transparency is particularly important in medicine. An understandable explanation of pose a major challenge. The data comes from different why the AI system recommends a particular sources, is often unstructured, and is prepared in dif- therapy or drug and what side effects can be ferent formats that use inconsistent terminology. That expected is critical. The real added value is why Dengel’s team, in cooperation with DFKI spin-off intelligentX GmbH, has built up extensive knowledge is only created when the physician under- stands the decisions.“ Prof. Andreas Dengel graphs that map hundreds of millions of links between genes, proteins, drugs, and diseases. AI searches these graphs and billions of cross-links to reveal unexpected patterns and connections. Explaining and understand- ing complex medical relationships achieves the neces- sary transparency, while discovering completely new associations. These large knowledge graphs are a core component of highly trained AI systems specialized for various tasks in diagnostics and drug development. Such graphs not only ensure accurate analyses, but also guarantee plausible results. This, in turn, generates trust and acceptance while enabling the fine-tuning of the models and further insights into medical processes and dependencies. Research meets application Dengel values close cooperation with partners from in- dustry and medical technology to ensure that the AI plat- form for biomedicine generates real world benefits. His methodology allows the development of technologies in a way that is precisely tailored to meet practical require- ments and quickly results in realistic applications. Such collaboration is an integral part of “ our work. It is the best way to conduct research that closely addresses the indus- try’s actual challenges, to gain new in- sights, and to effectively collaborate with experts from other disciplines to develop solutions.“ Prof. Andreas Dengel 8
AI Innovation Schematic representation of the knowledge graph: A team from DFKI and DFKI spin-off intelligentX GmbH has developed extensive knowledge graphs that link complex biomedical data, enabling AI-supported insights into previously hidden connections between genes, proteins, drugs, and diseases. This collaboration ensures that technological develop- in the healthcare sector. The complex approval process ments are precisely tailored to practical requirements, required for AI-supported medical devices represents an leading quickly to realistic applications. At the same time, additional hurdle. A clear, uniform set of rules would send it provides an ideal foundation for training young scien- a decisive signal for investment and faster development. tists to think and act in a technically sound and practical manner. Chance for personalized medicine AI is revolutionizing the ability to make detailed, cus- The TransferLabs at DFKI are an important vehicle for im- tomized therapies for an individual patient. Researchers plementing such close partnerships. These labs provide are able to analyze extensive biomedical data and, with a fertile training ground where talented young scientists new insights, to personalize drug development more can think and act in a manner that is both technically accurately. The technology promises not only better sound and application-oriented. Employees from partner treatments with fewer side effects, but more efficient organizations work directly with the DFKI research teams drug development overall. Initial successes can already on specialized cooperation platforms. Custom-made be seen in the discovery of new biomarkers and in the solutions and prototypes are developed and tested in the precise prediction of drug effects. The goal is personal- protected data room, while staying closely aligned with ized, transparent medicine that reaches many people and any scientific advances. The TransferLab, in effect, cre- significantly improves the quality of life. ates a dynamic exchange that accelerates innovation. Ethical and regulatory challenges At the intersection of medicine and artificial intelligence, laws and ethics are indispensable. When working with sensitive patient data, DFKI relies on state-of-the-art pri- vacy methods to protect information in such a way that it is no longer possible to identify individuals. There is also an important need to strike a balance between maximum data protection and rapidly advancing innovation. Incon- sistent regulation, such as in the European AI Act, cur- rently leads to uncertainty among companies, especially 9
AI Innovation PAIRS: Proactive crisis management with AI Companies that use artificial intelligence are not only more efficient, but also more resilient. Researchers in the PAIRS (Privacy-Aware, Intelligent and Resilient Crisis Management) project are developing a platform for efficient, data-based crisis management. The research team developed several AI-based smart crisis manage- ment services to identify crisis scenarios more quickly. The AI services provide decision-making bases for companies, governments, health organizations, and civil protection agencies, enabling them to proactively manage crises. Transparent supply chains In the domains of production, logistics, and supply Beyond the project: chains, organizations use the “Hidden Problem Detec- tor” module to identify hidden problems and optimize The Smart Service Engi- their component-based supply chains. The module neering research group led provides supply chain transparency using a graph- by Professor Maaß at DFKI theoretical approach to analyze bottleneck trends at in Saarbrücken develops the component level. This involves automatically con- solutions for the systemat- verting bill of materials (BOM) data into a knowledge ic design and development graph and enriching it with additional information, of smart service systems, such as price data. addressing the intersec- tion of technical and busi- ness management issues. Prof. Dr.-Ing. Wolfgang Maaß Head of Research Department Smart Service Engineering Secure power supply thanks to AI The “Outage Predictor” forecasts regional power ou- tages in the energy sector, thereby contributing to the resilience of the power supply. AI-based scenario planning in JSON-LD uses patterns based on past weather and power outage data to prepare a know- ledge graph showing potential crisis scenarios. This allows regional power outages to be predicted and ex- plained more accurately using the scenario patterns. 10
AI Innovation Comprehensive crisis and security management thanks to PAIRS An integrated, proactive crisis management is enabled by technology that covers various domains and features “Newspaper Signaling” modules) and reliably detecting oncoming earthquakes based on seismic sensor data (“REAVER” module). a modular architecture: The platform offers the addi- Thanks to its modular architecture and coverage of dif- tional options of identifying crisis signals in social me- ferent domains, the technology enables holistic, proac- dia and news at an early stage (“Social Signaling” and tive crisis management. 11
AI Innovation Carve-DL: AI is revolutionizing digital forensics Precise data reconstruction for more security Modern forensics relies on digital clues. Criminals, without realizing it, often leave traces behind – for example, when using computers or the internet. To secure such digital evidence, investigators rely on state-of-the-art technologies – including specially trained “data sniffing” dogs that can even detect hidden USB sticks. But what can be done when essential data has been irretrievably deleted? At DFKI, Carve-DL is providing answers and revolutionizing digital forensics. Rethinking digital evidence In the digital age, standard investigations quickly reach their limits when analyzing large data volumes if data is only available in disjointed fragments, often in unlinked storage areas. Fragmenta- tion makes traditional data reconstruction methods considerably more difficult. Carve-DL applies the latest « We see Carve-DL and other AI projects as a sustainable approach to obtaining relevant evidence more quickly – an important building block for efficient and modern police work. » Dr. Matthias Müller, LKA Rhineland-Palatinate methods. The evidence un- covered by Carve-DL is used to make investigations faster, more reliable, and more accu- rate. Practical uses The technology not only helps with police and law enforce- ment investigations, but also supports companies with data recovery and the protection of digital archives. Carve-DL deep learning technologies, which include transformer works closely with security agencies to ensure that re- models that are able to precisely reconstruct even small search and application go hand in hand. In other words, pieces of data to recover information previously consid- innovation applied where it is needed. ered lost. AI brings order to data chaos Neural transformer models and intelligent analysis Outlook Carve-DL represents a key technology for the digital se- curity of tomorrow. It takes forensics and data recovery methods, such as iterative clustering turn, scattered bits to a whole new level, setting new standards as it trans- of data back into usable evidence. These digital traces forms research and society. are virtually impossible to detect using conventional 12
New classification models for identifying file types in raw data improve the recovery process, supplemented by deep learning-based clustering techniques that efficiently identify related file fragment groups. AI Innovation Data- Recovery Delete Disk 1 Fragmentation Overview 3 An advanced fragment reordering model that already correctly assembles 95% of the reconstructed image fragments Lorem ipsum dolor sit amet, consecte Disk 1 Reordering Recovered data 95% 2 A specialized verification model to reliably reconstruct image fragments. CARVE-DL Carve-DL was developed in collabora- tion with the Rhineland-Palatinate State Criminal Police Office and the Federal Criminal Police Office. As part of the GeT- AI working group led by Dr. Tobias Wirth, DFKI researchers are investigating how modern AI tools can efficiently relieve investigative authorities of their increas- ingly complex tasks and contribute to improving crime fighting and detection in the long term. Disk 1 Data-Recovery 1 New classification models to identify file types in raw data, which improve the recovery process File type Sound Video Image Text Delete process Reconstruction process
AI Innovation A network of possibilities: Into the networked future with Open6GHub+ DFKI is transforming its research hub into an open transfer center for the next genera- tion of mobile communications. The aim of Open6GHub+ is to make joint research with institutes and universities widely available to companies, other research institutions, and society at large. Development is extremely complex as different net- Industry and society works have traditionally operated independently of one To reach the full potential of 6G technology, all devel- another — with different frequency bands, systems, and opments at Open6GHub+ are tested in a practical en- regulatory requirements. Open6GHub+ brings these vironment. Together with industry partners, technical various systems together in one ecosystem, creating a systems are jointly tested in open labs in so-called reliable, extensive, and disaster-proof communications “missions” and applications are readied for market en- infrastructure. try from 2030 onwards. The latest generation of mobile communications offers major innovations: these include the integration of intelligent services, innovative net- work architectures, and the close interconnection of « We have a wide range of 6G terrestrial and non-terrestrial networks for use with sat- technologies that we are deve- ellites and drones. The aim is to create a more robust, loping with partners from in- more resilient infrastructure, specifically addressing the dustry for transfer to concrete weaknesses that 5G leaves unresolved, to guarantee applications. » European digital sovereignty. Prof. Dr.-Ing. Hans Dieter Schotten Head of Intelligent Networks Research, Coordinator of the 6G-Plattform, and Head of Open6GHub+ Many of the current generation’s promises such as high throughput in campus data networks or precision locali- zation have not been fully realized. Many companies and organizations still continue to rely on the proven, older, or alternative systems. Open6GHub+ is closing the gaps with 6G while creating more tangible benefits for the economy and society. 14
AI Innovation To realistically map the development specifications for ness, and the public. Specifically, this calls for educa- 6G, Open6GHub+ applies two main concepts: tion, transparent communication, and opportunities for 1. Missions: Technological goals are specified and im- OpenLabs and the successful missions, Open6GHub+ is plemented in collaboration with partner organizations making an important contribution to meeting these re- participation. Through the effective implementation of from industry and society at large. The resulting techno- quirements. logical output flows directly into international standards and serves as the basis for new products. The next generation of mobile communications will not only strengthen the economy, but above all, enrich 2. OpenLabs: These open laboratories are places for people’s lives. Secure, reliable, and comprehensive net- joint development, demonstration, and validation of works pave the way for digital participation for all. technologies together with partners from industry and society. SMEs and start-ups also have access and oppor- tunities to contribute here. AI as the heart of 6G Artificial intelligence has penetrated all key areas of so- ciety and the economy. A modern 6G network creates the foundation for all individuals and organizations to access various AI application technologies. AI makes real-time orchestration of increasingly complex network structures possible, which contributes to more efficient and sustainable use of resources. 6G and AI form a symbiotic unit that benefits all users 6G Platform and Innovation Space The 6G Platform coordinates accompanying re- search for all 6G projects funded by the Feder- al Ministry of Research, Technology, and Space (BMFTR). Since its establishment in 2021, it has made a significant contribution to Germany’s equally: distributed AI applications running on edge digital sovereignty. devices in factories, hospitals, vehicles, or in everyday private life as well as intelligent services provided in- dependently through the network. This interaction pro- vides the technological foundation for continuously learning systems – whether care robots, autonomous vehicles, smart environments, or individual assistance systems and serves to flexibly adapt to the needs of all users. Innovation comes with challenges The development of 5G taught us that the successful rollout of 6G requires the involvement of politics, busi- The DFKI consolidates its excellent research expertise around Open6GHub – the current research hub – its new transfer center, Open- 6GHub+, and the start.smart.connect startup incubator, which specifically supports and gui- des young companies during the critical start- up phase. This dense network makes DFKI one of the central hotspots for the development and implementation of 6G technologies in Germany. 15
AI Innovation ESCADE: Sustainable and efficient operation of data centers courtesy of AI Artificial intelligence has huge potential, but the power consumption by this technology is huge: according to an International Energy Agency (IEA) forecast, the global demand for electricity to power the data centers used to train and operate AI models will more than double by 2030. This is precisely where the ESCADE project of DFKI and Saarland University comes in. The goal is to improve the energy efficiency of AI models by up to 90 percent. Achieving this goal will enable companies to reach their sustainability goals without compromising performance. Generative AI: powerful but energy-intensive Large amounts of data are required for natural lan- them without compromising performance. If all goes well, the advantages of AI will be accessi- guage processing AI models, such as ChatGPT, Mistral ble to small and medium-sized enterprises as well. and Claude, and for visual AI models, to reach their full potential. Generative AI models therefore consume a lot of electricity. Even training the models is extreme- Smaller AI models, optimized performance The ESCADE project focuses on compressing AI models. ly resource-intensive: some estimates suggest that One way this is achieved is through knowledge distilla- training GPT-4 consumed 50 GWh, and ongoing opera- tion. Here, smaller and resource-saving “student mo- tion even demands more energy. According to Sam Altman, ChatGPT needs about 0.34 watt hours to process each query, and OpenAI reports that there are 2.5 billion que- ries per day. The ESCA- DE project aims to redu- ce the size of AI models so companies can use 16 dels” are derived from large “teacher models.” These do not contain all the data, but only that which is relevant to the specific task. The pro- ject results show that this approach can reduce the size of the models by up to 90 percent. The researchers also use the method of “neu- ral architecture search.”
AI Innovation different AI model configurations and hardware options on performance. This is where various efficiency-increa- sing strategies can be applied. EAVE supports informed decisions that balance the performance, sustainability, and cost-effectiveness of data centers. Decision-makers can thus manage their data centers in a data-driven, re- source-efficient, and future-proof manner. In addition, companies can specifically optimize the ratio of AI per- formance to operating costs. ESCADE at Saarstahl: Specialized AI model provides better sustainability and efficiency In cooperation with Stahl Holding Saar, the project team from the Smart Service Engineering research area led by Professor Wolfgang Maaß is testing the practical applicability of the research. The aim is to combat scrap shortages by sorting scrap steel so that it can be used to produce new steel. Until now, the company has used a large visual model for scrap sorting. Researchers found a way to make this more compact. As a result, the visual model is not only more energy efficient, but also, in some cases, more power- ful. Instead of the large all-rounder model, which con- sumes a lot of energy, the task is now performed by a small, specialized AI model. While neural networks for AI models are normally desig- ned manually by experts, in this case the AI itself takes on this task: it automatically identifies the optimal archi- tecture for the respective application. The researchers have already achieved promising results, especially with visual AI models, i.e., those that process digital image data. Current project results show that models can also be reduced in size by up to 90 percent in this way. Hardware updates In addition to software optimization, the researchers are also focusing on hardware. They are developing con- cepts for integrating neural processing units (NPUs) into GPU-based data centers. This involves the use of neuro- morphic chip technologies. Unlike traditional AI chips, these are based on the structure of the human brain and the way it processes information. This results in chips that enable AI models to run efficiently and with high performance. One tool, full transparency: EAVE With the prototype “EAVE – Energy Analytics for Cost- Effective & Sustainable Operations,” researchers have developed a tool that comprehensively monitors data centers and enables efficient and sustainable operation. EAVE forecasts the energy consumption, CO₂ emissions, and operating costs of an AI model and provides an over- view of the country and time at which the models can be executed in the most energy- and cost-efficient manner. At the same time, the system illustrates the impact of 17
Interview Artificial Intelligence – The engine of transformation Double interview with Prof. Antonio Krüger, CEO DFKI, and Prof. Andreas Dengel, Executive Director at DFKI Kaiserslautern In which areas of the economy and society is artificial intelligence currently having the greatest transformative impact – and what res- ponsibilities does this entail? patterns of thinking. The potential root cause of an issue. That is why of this technology can change en- a key element of our work is to tire industries, societies, and life’s establish close contact with users. realities for all of us. At DFKI, this Such contact enables us to develop means that we must use this power meaningful research projects – responsibly and – together with scientifically relevant, practical, and partners from science, industry, and applicable. society – shape it at an early stage. What is the potential of AI as a catalyst for social change and the creation of new value, and how can this be translated into concrete innovations? Krüger: The full potential of AI becomes evident where it addresses soci- etal issues. Whether in healthcare, manufacturing, or administration, AI opens key sectors for the future, for example, the accelerated de- Prof. Antonio Krüger, Prof. Andreas Dengel, velopment of new therapies. It is CEO DFKI: Executive Director DFKI: more than a tool – it is the engine AI is the cross-cutting technolo- Real innovation can only occur of innovative business models and gy of our time, comparable to the when dealing with real problems. social progress. wheel, the steam engine and elec- We need to understand the con- tricity. It is revolutionizing process- text of why or how problems come Dengel: es and opening up completely new about before we can get to the An impressive example is the repur- 18
Interview posing of approved drugs, where research alone is not enough. The Krüger: the analysis of clinical data, mo- impact comes about only when in- DFKI is a major driver of innovation lecular profiles, and publications is novation is embedded into a social – from robots that work closely with enabling the identification of new context. indications. NLP extracts data from unstructured texts, machine learn- Dengel: humans to AI systems in healthcare that produce ground-breaking therapeutic success in cases of ing recognizes hidden patterns, and Successful transfer requires joint severe and rare diseases. The simulations test drug-protein inter- projects and visible successes – measure of our success is how well actions. This is where innovation for example, as in our transfer labs we use AI to strengthen and and social benefit come together. where we develop AI-supported transform the society and the In your opinion, what is needed for AI innovations to have their full impact in practical and economic transfer? solutions for observing cell cul- economy – by innovative approach- tures, modernizing forensics, and es and trusted partnerships. ensuring security in the financial world. Dengel: The measurable results of many exchange of know-how and best collaborations strengthen Germa- practices across national borders What’s more, we promote the Krüger: ny’s position as a leader in innova- and support the training of skilled We create trust and innovation by tion. building bridges between cutting- edge research and business. We transform scientific excellence into tangible products and solutions by partnering in over 400 projects in industry and society. However, In what ways does DFKI influence economic and social change in its capacity as a research and transfer partner? workers in the field of AI. In this way, we are strengthening Europe’s long-term innovative power, ensuring the sustainability of technological progress, and creat- ing opportunities for the societies of tomorrow. Learn more about how we create real added value with our partners: 19
Lübeck Oldenburg Bremen Berlin Osnabrück DFKI Site DFKI Laboratory DFKI Branch Office Trier Darmstadt Kaiserslautern Saarbrücken Imprint dfki ai next; Issue November | 2025; Published by: Deutsches Forschungszentrum für Künstliche Intelligenz GmbH (DFKI); Trippstadter Str. 122, D-67663 Kaiserslautern; Phone: +49 631 20575 0; E-Mail: news@dfki.de; Editors: Jeremy Gob (responsible), Valentina Röther, Christian Heyer; Layout: Lando Lehmann, Juan Mejia; Production: One Vision Design, Saarbrücken Credits Title: DFKI; Page 3: DFKI, Jürgen Mai; Page 4: ESA/ATG medialab; Page 5 top: DFKI, right: DFKI, Juan Mejia; Page 6: DFKI, Lando Lehmann; Page 7: DFKI, Felix Amsel; Page 8: DFKI; Page 9: top: DFKI, Lando Lehmann; bottom: DFKI; Page 10: DFKI, Jürgen Mai; Page 11: ai-generated image, Adobe Firefly, Mahmud – stock.adobe.com; Page 13: DFKI, Juan Mejia; Page 14: DFKI, Lando Lehmann; Page 16: SaymaHossan – stock.adobe.com; Page 17: diloomi – stock.adobe.com; Page 18: DFKI, Felix Amsel; DFKI, Jürgen Mai; Page 19: stocked – stock.adobe.com; Page 20: DFKI