Showing papers in "Künstliche Intelligenz in 2014"

Strategic Argumentation in Multi-Agent Systems

Abstract: Argumentation-based negotiation describes the process of decision-making in multi-agent systems through the exchange of arguments. If agents only have partial knowledge about the subject of a dialogue strategic argumentation can be used to exploit weaknesses in the argumentation of other agents and thus to persuade other agents of a specific opinion and reach a certain outcome. This paper gives an overview of the field of strategic argumentation and surveys recent works and developments. We provide a general discussion of the problem of strategic argumentation in multi-agent settings and discuss approaches to strategic argumentation, in particular strategies based on opponent models.

A Survey of Multi-Agent Decision Making

Abstract: In this article we give a high-level overview of various aspects relevant to multi-agent decision making. Classical decision theory makes the start. Then, we introduce multi-agent decision making, focussing on game theory, complex decision making, and on intelligent agents. Afterwards, we discuss methods for reaching agreements interactively, e.g. by negotiation, bargaining, and argumentation, followed by approaches to coordinate and to control agents’ decision making.

The RACE Project : Robustness by Autonomous Competence Enhancement

Abstract: This paper reports on the aims, the approach, and the results of the European project RACE. The project aim was to enhance the behavior of an autonomous robot by having the robot learn from conceptualized experiences of previous performance, based on initial models of the domain and its own actions in it. This paper introduces the general system architecture; it then sketches some results in detail regarding hybrid reasoning and planning used in RACE, and instances of learning from the experiences of real robot task execution. Enhancement of robot competence is operationalized in terms of performance quality and description length of the robot instructions, and such enhancement is shown to result from the RACE system.

Towards Learning of Generic Skills for Robotic Manipulation

Abstract: Learning versatile, reusable skills is one of the key prerequisites for autonomous robots. Imitation and reinforcement learning are among the most prominent approaches for learning basic robotic skills. However, the learned skills are often very specific and cannot be reused in different but related tasks. In the project "Behaviors for Mobile Manipulation", we develop hierarchical and transfer learning methods which allow a robot to learn a repertoire of versatile skills that can be reused in different situations. The development of new methods is closely integrated with the analysis of complex human behavior.

Rough Terrain 3D Mapping and Navigation Using a Continuously Rotating 2D Laser Scanner

Abstract: Mapping, real-time localization, and path planning are prerequisites for autonomous robot navigation. These functions also facilitate situation awareness of remote operators. In this paper, we propose methods for efficient 3D mapping and real-time 6D pose tracking of autonomous robots using a continuously rotating 2D laser scanner. We have developed our approach in the context of the DLR SpaceBot Cup robotics challenge. Multi-resolution surfel representations allow for compact maps and efficient registration of local maps. Real-time pose tracking is performed by a particle filter observing individual laser scan lines. Terrain drivability is assessed within a global environment map and used for planning feasible paths. Our approach is evaluated using challenging real environments.

Technologies for the Fast Set-Up of Automated Assembly Processes

Abstract: In this article, we describe technologies facilitating the set-up of automated assembly solutions which have been developed in the context of the IntellAct project (2011–2014). Tedious procedures are currently still required to establish such robot solutions. This hinders especially the automation of so called few-of-a-kind production. Therefore, most production of this kind is done manually and thus often performed in low-wage countries. In the IntellAct project, we have developed a set of methods which facilitate the set-up of a complex automatic assembly process, and here we present our work on tele-operation, dexterous grasping, pose estimation and learning of control strategies. The prototype developed in IntellAct is at a TRL4 (corresponding to ‘demonstration in lab environment’).

Measuring Inconsistency in Multi-Agent Systems

Abstract: We introduce and investigate formal quantitative measures of inconsistency between the beliefs of agents in multi-agent systems. We start by recalling a well-known model of belief in multi-agent systems, and then, using this model, present two classes of inconsistency metrics. First, we consider metrics that attempt to characterise the overall degree of inconsistency of a multi-agent system in a single numeric value, where inconsistency is considered to be individuals within the system having contradictory beliefs. While this metric is useful as a high-level indicator of the degree of inconsistency between the beliefs of members of a multi-agent system, it is of limited value for understanding the structure of inconsistency in a system: it gives no indication of the sources of inconsistency. We therefore introduce metrics that quantify for a given individual the extent to which that individual is in conflict with other members of the society. These metrics are based on power indices, which were developed within the cooperative game theory community in order to understand the power that individuals wield in cooperative settings.

Autonomous Navigation for On-Orbit Servicing

Abstract: On-orbit servicing missions induce challenges for the rendezvous and docking system since a typical target satellite is not specially prepared for such a mission, can be partly damaged or even freely tumbling with lost attitude control In contrast to manned spaceflight or formation flying missions, new sensors and algorithms have to be designed for relative navigation Dependent on the distance to the target, optical sensors such as mono and stereo cameras as well as 3D sensors like laser scanners can be employed as rendezvous sensors Navigation methods for far and close range and different verification methods are discussed

Mental Models for Intelligent Systems: eRobotics Enables New Approaches to Simulation-Based AI

Abstract: eRobotics is a newly evolving branch of e-Systems engineering, providing tools to support the whole life cycle of robotic applications by means of electronic media. With the eRobotics methodology, the target system and its environment can be modeled, validated, and calibrated to achieve a close-to-reality simulation. In this contribution, we present simulation-based mental models for autonomous systems as a foundation for new approaches to prediction and artificial intelligence. We formulate a methodology to construct optimization problems within simulation environments in order to assist autonomous systems in action planning. We illustrate the usefulness and performance of this approach through various examples in different fields. As application for space robotics, we focus on climbing strategies of a legged mobile exploration robot. Furthermore, we enable skillfull interaction control in service robotics and address energy consumption issues. The contribution concludes with a detailed discussion of the concept presented here.

Responsible Intelligent Systems

Abstract: The 2013 ERC-consolidator project “Responsible Intelligent Systems” proposes to develop a formal framework for automating responsibility, liability and risk checking for intelligent systems. The goal is to answer three central questions, corresponding to three sub-projects of the proposal: (1) What are suitable formal logical representation formalisms for knowledge of agentive responsibility in action, interaction and joint action? (2) How can we formally reason about the evaluation of grades of responsibility and risks relative to normative systems? (3) How can we perform computational checks of responsibilities in complex intelligent systems interacting with human agents? To answer the first two questions, we will design logical specification languages for collective responsibilities and for probability-based graded responsibilities, relative to normative systems. To answer the third question, we will design suitable translations to related logical formalisms, for which optimised model checkers and theorem provers exist. All three answers will contribute to the central goal of the project as a whole: designing the blueprints for a formal responsibility checking system. To reach that goal the project will combine insights from three disciplines: philosophy, legal theory and computer science.

Development, Control, and Empirical Evaluation of the Six-Legged Robot SpaceClimber Designed for Extraterrestrial Crater Exploration

Abstract: In the recent past, mobile robots played an important role in the field of extraterrestrial surface exploration. Unfortunately, the currently available space exploration rovers do not provide the necessary mobility to reach scientifically interesting places in rough and steep terrain like boulder fields and craters. Multi-legged robots have proven to be a good solution to provide high mobility in unstructured environments. However, space missions place high demands on the system design, control, and performance which are hard to fulfill with such kinematically complex systems. My dissertation focuses on the development, control, and evaluation of a six-legged robot for the purpose of lunar crater exploration considering the requirements arising from the envisaged mission scenario. The performance of the developed system is evaluated and optimized based on empirical data acquired in significant and reproducible experiments performed in a laboratory environment in order to show the capability of the system to perform such a task and to provide a basis for the comparability with other mobile robotic solutions.

Smart Grid Challenges for Electricity Retailers

Abstract: The need to achieve sustainability is driving a major transformation of the energy sector. The traditional top-down approach to electricity supply and grid management is being strongly disrupted by a range of forces including distributed renewables, retail market liberalization, and the need for energy consumers to adapt their behavior to the availability of renewable energy sources. We introduce Power TAC, a competitive simulation that challenges researchers to build autonomous trading agents that tackle the complex decision processes a retailer will need to face in future competitive retail electricity markets.

Beyond Reinforcement Learning and Local View in Multiagent Systems

Abstract: Learning is an important component of an agent’s decision making process. Despite many messages in contrary, the fact is that, currently, in the multiagent community it is mostly likely that learning means reinforcement learning. Given this background, this paper has two aims: to revisit the “old days” motivations for multiagent learning, and to describe some of the work addressing the frontiers of multiagent systems and machine learning. The intention of the latter task is to try to motivate people to address the issues that are involved in the application of techniques from multiagent systems in machine learning and vice-versa.

Regularization-Based Multitask Learning With Applications to Genome Biology and Biological Imaging

Abstract: The aim of multitask learning is to improve the generalization performance of a set of related tasks by exploiting complementary information about the tasks. In this paper, we review established approaches for regularization based multitask learning, sketch some recent developments, and demonstrate their applications in Computational Biology and Biological Imaging.

Advanced Driver Assistance Systems and Animals

Abstract: Advanced driver assistance systems are widely used. Some support and inform the driver. Others relieve him or her of certain tasks—and transform the human-guided system into a semi-autonomous one. For some years also fully autonomous systems have been on the roads, so-called self-driving cars, as prototypes of companies and within research projects. From the perspective of ethics—both of the special fields of ethics like animal ethics, information ethics and technology ethics and of machine ethics which can be understood as a counterpart to human ethics—advanced driver assistance systems raise various questions. The aim of this paper is to derive suggestions from animal ethics and other disciplines for the improvement and development of the systems. The basis are literature analysis and own classifications and considerations. The result is that there are many possibilities to expand existing systems and to develop new functions in the context with the aim to reduce the number of animal victims.

Empirical Evidence for Context-aware Interfaces to Pedestrian Navigation Systems

Abstract: For geographical mobile search tasks it is rarely sufficient to assist users identifying what location they are currently looking for, e.g. a store, cafe or museum. Often the user needs support in being guided to a retrieved location in a physical space. This means that mobile search is strongly connected with navigation. There is a large body of work indicating that navigating towards points of interest is challenging for many people. In this work we explore how to support best this part of the task by investigating how objects in the physical world—landmarks—can be used in information systems to guide people to their desired location. We present the results of a series of eye tracking studies on the orientation behavior of persons executing indoor navigation tasks. The main finding of the studies is that the contextual relevance and the function of a landmark for completing the task efficiently matters more than the context-free salience of the same landmark. The findings have implications for the design of mobile search systems that support geographical search tasks as they lead to new context-adaptive strategies for navigation systems to explain routes. We provide evidence that even the interface has to adapt its content on the state of the navigation task and the current spatial context in order to provide user- and context-adaptive intuitive interaction.

Special Issue on Transfer Learning

Abstract: Machine learning (ML) has been recognized as central to artificial intelligence (AI) for many decades, if not from the early beginnings of AI. The question of how the things that have been learned in one context can be re-used and adapted in related contexts, however, has only been brought to the attention of the wider ML research community over the past few years. In parallel (and sometimes preceding this), transfer learning has been receiving increasing attention in other research areas, e.g. psychology. In this special issue, we present a snapshot of different perspectives on the issues surrounding transfer learning, that we hope will inspire readers to further exploration of this exciting and challenging resarch field. The contributions to this special issue are a reflection of the diversity of the research area of transfer learning (TL) and the wide range of application domains. In the first article, Haitham Bou Ammar et al. present a discussion of TL for reinforcement learning (RL). RL has been a focus of much TL research due to its suitability for autonomous intelligent agents. Bou Ammar et al. suggest that some of the tasks that needed to be carried out by the designer, such as state feature mapping, can be automated. This discussion article is followed by three papers presenting various applications of TL, namely robotic manipulation (Jan Hendrik Metzen et al.), automated negotiation (Siqi Chen et al.), and computational biology (Christian Widmer et al.). We are also happy to include in this issue an interview with Peter Stone and Matthew Taylor, both prolific RL researchers and amongst the pioneers of TL for RL. The special issue is rounded off by two research reports. Michael Siebers reports on a project applying TL to plan generation. Marco Ragni and Gerhard Strube look at TL from the cognitive psychology viewpoint, discussing insights into human reasoning and what makes transfer problems difficult to solve for humans. I would like to thank Ute Schmid for her invaluable help in putting this issue together, and the authors for their contributions.

Space Robotics: An Overview of Challenges, Applications and Technologies

Abstract: While space exploration may be considered anything but dull, it certainly is very dangerous. Expanding our knowledge on the solar system to look for clues to such fundamental questions as the origins of life, or a sustained human presence on anything other than earth may well be worth the risk. The involved costs for mitigating the risk of human space flight are prohibitive. Robotic missions, like the hugely successful Mars Exploration Rovers, have shown that robotics as a sub-field of Artificial Intelligence can perform scientific exploration activities without human presence, and will play an even more prominent role in future mission scenarios. Worldwide technology research efforts are continuously expanding the capabilities of mobile robotic systems. This article provides an overview of the special conditions and examples of technological solutions for the development of space robots, as well as different fields of application.

Two-Phased Knowledge Formalisation for Hydrometallurgical Gold Ore Process Recommendation and Validation

Abstract: This paper describes an approach to externalising and formalising expert knowledge involved in the design and evaluation of hydrometallurgical process chains for gold ore treatment. The objective was to create a case-based reasoning application for recommending and validating a treatment process of gold ores. We describe a twofold approach. Formalising human expert knowledge about gold mining situations enables the retrieval of similar mining contexts and respective process chains, based on prospection data gathered from a potential gold mining site. Secondly, empirical knowledge on hydrometallurgical treatments is formalised. This enabled us to evaluate and, where needed, redesign the process chain that was recommended by the first aspect of our approach. The main problems with formalisation of knowledge in the domain of gold ore refinement are the diversity and the amount of parameters used in literature and by experts to describe a mining context. We demonstrate how similarity knowledge was used to formalise literature knowledge. The evaluation of data gathered from experiments with an initial prototype workflow recommender, Auric Adviser, provides promising results.

Cognitive Complexity and Analogies in Transfer Learning

Abstract: The ability to learn often requires transferring relational knowledge from one domain to another. It is difficult for humans and computers to identify the respective source domain from which relational characteristics can be applied to the target domain. An additional source of human reasoning difficulty is the complexity of the transformation function. In this article we investigate two domains in which the identification of relational patterns and of a transformation function are necessary: number series and geometrical analogy problems. Characteristics of the human processes are presented and existing cognitive models are discussed.

Characterisation of Large Changes in Wind Power for the Day-Ahead Market Using a Fuzzy Logic Approach

Abstract: Wind power has become one of the renewable resources with a major growth in the electricity market. However, due to its inherent variability, forecasting techniques are necessary for the optimum scheduling of the electric grid, specially during ramp events. These large changes in wind power may not be captured by wind power point forecasts even with very high resolution numerical weather prediction models. In this paper, a fuzzy approach for wind power ramp characterisation is presented. The main benefit of this technique is that it avoids the binary definition of ramp event, allowing to identify changes in power output that can potentially turn into ramp events when the total percentage of change to be considered a ramp event is not met. To study the application of this technique, wind power forecasts were obtained and their corresponding error estimated using genetic programming and quantile regression forests. The error distributions were incorporated into the characterisation process, which according to the results, improve significantly the ramp capture. Results are presented using colour maps, which provide a useful way to interpret the characteristics of the ramp events.

Is Model-Based Robot Programming a Mirage? A Brief Survey of AI Reasoning in Robotics

Abstract: Researchers in AI and Robotics have in common the desire to “make robots intelligent”, evidence of which can be traced back to the earliest AI systems. One major contribution of AI to Robotics is the model-centered approach, whereby intelligence is the result of reasoning in models of the world which can be changed to suit different environments, physical capabilities, and tasks. Dually, robots have contributed to the formulation and resolution of challenging issues in AI, and are constantly eroding the modeling abstractions underlying AI problem solving techniques. Forty-eight years after the first AI-driven robot, this article provides an updated perspective on the successes and challenges which lie at the intersection of AI and Robotics.

DLR SpaceBot Cup 2013: A Space Robotics Competition

Abstract: In November 2013, the German Aerospace Center (DLR) in Bonn hosted the SpaceBot Cup, Germany’s first of its kind space robotics competition The scenario is set in a planetary exploration environment with some manipulation tasks Ten entrants had eight month to define, develop, and build robotic systems and the according ground station setup to conduct a remote testbed mission Then, the robotic element(s) were deployed onto a sparsely known planetary surface and had to conduct exploration of the environment, find and collect two artificial objects, and mount them to a third object Communication between ground control station and planetary surface was limited and impaired by delay, making autonomous functionality crucial for the success of the mission In this report, the motivation, scenario, tasks, and final competition event of the DLR SpaceBot Cup 2013 are presented

Modeling Reality for Camera Registration in Augmented Reality Applications

Abstract: One of the central problems of Augmented Reality is to make reality and virtual objects coincide in some way. Any technique aiming at solving this problem requires an internal way of representing the real objects of interest, i.e. the objects that the system expects to see. We name these representations Reality Models and take in this work a closer look at the various ways of representing reality in the context of AR. We propose a classification of AR applications based on the requirements on Reality Models. AR applications can be first classified into applications where the full 6DOF camera pose is recovered and applications where a 2D object localization in the image space is sufficient. We further detail the classification by providing examples of planar fiducials and textured objects in the first case and object detection and local pose estimation in the second case. In all provided examples we extend the state of the art by providing new Reality Models or better ways to construct and use existing ones. Augmented Reality is a powerful and intuitive technology that can be used in various situations, ranging from small smartphone games to large-scale outdoor augmentations for marketing and tourism. However, these diverse possibilities share one common feature: in order to augment reality, at least some parts of the real environment need to be known in advance and modeled in an appropriate way. This prior knowledge is necessary to allow an AR system to build the ‘‘link’’ between the real world and its virtual counterpart. In this context, the challenge is that the mathematical models used for real objects have to meet the requirements of an AR application: the representation should be sparse to ensure a low memory footprint, robust to different kinds of alteration that arises in the context of optical sensors (illumination changes, partial occlusions, shape deformation) and at the same time complete enough to allow for recovering 3D information (object position or complete camera pose) in real time. A look at the state of the art shows that this aspect of Augmented Reality has never been analyzed thoroughly. In this work, our goal is to investigate how real objects and scenes can be efficiently modeled for AR applications and to develop solutions for reality modeling and pose estimation for different types of AR scenarios. We therefore propose a classification of Reality Models based on the type of real objects used, the nature of the virtual augmentation and the type of registration they permit. In this context, we present novel representations, and provide a detailed analysis of their use in AR. Depending on the type of information available on the real objects present in the scene, and on the type of AR application being built, two different approaches for registration can be developed: a full 3D registration of the camera or a 2D object labeling approach (see Fig. 1). A full 3D registration means that all the parameters of the camera are known. As a consequence, virtual objects can be inserted in the real scene at an exact position and they appear as if they were completely integrated in the environment. Reality Models for this approach can be further divided into markerbased models, and models based on textured objects. 2D object labeling is used when the focus of the application is not the integration of virtual objects in a real scene, but rather the automatic identification of objects or scene parts in order to provide contextual information to the user. For this approach, the Reality Models can use object detection or extended object detection (Fig. 1). A. Pagani (&) DFKI GmbH, Trippstadter Straße 122, 67655 Kaiserslautern, Germany e-mail: alain.pagani@dfki.de

The Virtual Space Robotics Testbed: Comprehensive Means for the Development and Evaluation of Components for Robotic Exploration Missions

Abstract: Designing robotic components and algorithms for exploration and science missions in space and on celestial bodies requires extensive preparation and testing. Due to the extremely high costs of spaceflight and space-capable hardware, design and evaluation iterations need to be performed on earth, using computer simulations as well as real test environments to make predictions about the robot’s performance on the real mission. A variety of ways to simulate single aspects of a robot like mechanical stress or power consumption in great detail are in active use. However, no truly holistic solution that evaluates overall system behavior across a multitude of disciplines has been developed to date. In this paper, we discuss the benefits of such a solution and present our virtual space robotics testbed, a framework that represents our work towards a holistic solution.

Transfer for Automated Negotiation

Abstract: Learning in automated negotiation is a difficult problem because the target function is hidden and the available experience for learning is rather limited. Transfer learning is a branch of machine learning research concerned with the reuse of previously acquired knowledge in new learning tasks, for example, in order to reduce the amount of learning experience required to attain a certain level of performance. This paper proposes a novel strategy based on a variation of TrAdaBoost—a classic instance transfer technique—that can be used in a multi-issue negotiation setting. The experimental results show that the proposed method is effective in a variety of application domains against the state-of-the-art negotiating agents.

Visual Learning of Semantic Concepts in Social Multimedia

Abstract: Currently, traditional media is experiencing a major shift towards social media. At the same time, interaction via social media is to an increasing degree enriched with images and videos, as seen during the Arab Spring in the Middle East in 2012 or the Boston Marathon Bombings on April 15, 2013. This combination gives rise to a new type of content, which is being called social multimedia. One key trigger for this trend is the capability to upload and distribute images and videos over the Internet minutes after such incidents happen on video-sharing platforms like YouTube or Vimeo. This is possible due to the availability of broadband Internet, the low price of storage, and the omnipresence of camera-equipped mobile devices. It allows people to record, publish, and share digital images and videos without notable effort. This ubiquity of visual content conveys much about our thinking and feeling, in that it reflects our personal lives and ourselves as a society. Unfortunately, this content is of little use if it is not accessible to users, e.g., by allowing users to retrieve videos by keyword-based search. However, keyword-based search requires each individual video to be annotated with a set of keywords describing its content. Given the vast amount of video content being created nowadays (YouTube, for example, stores about 100 hours of video content every minute) this poses an impossible task for human annotators. Worse, as naturally as humans can perceive their surroundings visually, this undertaking is quite challenging for machines. This lack of correspondence between the lowlevel features that machines can extract from videos (i.e., the raw pixel values) and the high-level conceptual interpretation a human associates with perceived visual content is referred to as the semantic gap [6]. In recent years, great effort has been spent on contentbased methods directly analyzing the video stream to bridge this gap. Following this line of research, the thesis focuses on concept detection [7], the task to detect semantic concepts in visual content. Given an input video clip, concept detection systems use statistical learning to infer the presence of a target concept by calculating its probability of appearance from low-level features extracted from the content. For this purpose, the set of all concepts— or concept vocabulary—should cover a broad spectrum of entities, such as objects (‘‘chair’’, ‘‘telephone’’), scene types (‘‘cityscape’’, ‘‘desert’’), and activities (‘‘interview’’, ‘‘people singing’’), requiring concept detection systems to provide detectors for hundreds or even thousands of target concepts. This, however, is considered as a major challenge in concept detection, as it demands labeled training samples for supervised machine learning—the underlying technology of current systems [7]. Such ground-truth training samples are usually acquired manually, i.e., a human annotator labels videos for whether the concept occurs. This time-consuming and cost-intensive effort creates a scalability problem, leading to smallscale, fixed concept vocabularies being useful in research setups, but making it impossible to satisfy the changing demands of users’ information needs. This leads to stateof-the-art systems still focusing on generic concepts such as ‘‘quadruped’’ or ‘‘hand’’ instead of providing detectors for concepts of interest, e.g., sports events such as ‘‘Olympics 2012’’, incidents such as the ‘‘Costa D. Borth (&) International Computer Science Institute & UC Berkeley, 1947 Center Street, Ste. 600, Berkeley, CA 94704, USA e-mail: borth@icsi.berkeley.edu

Automated Transfer for Reinforcement Learning Tasks

Abstract: Reinforcement learning applications are hampered by the tabula rasa approach taken by existing techniques. Transfer for reinforcement learning tackles this problem by enabling the reuse of previously learned behaviours. To be fully autonomous a transfer agent has to: (1) automatically choose a relevant source task(s) for a given target, (2) learn about the relation between the tasks, and (3) effectively and efficiently transfer between tasks. Currently, most transfer frameworks require substantial human intervention in at least one of the previous three steps. This discussion paper aims at: (1) positioning various knowledge re-use algorithms as forms of transfer, and (2) arguing the validity and possibility of autonomous transfer by detailing potential solutions to the above three steps.

Using Presence, Spatial Formations and Sociometry to Measure Interaction Quality in Mobile Robotic Telepresence Systems

Abstract: The use of video mediated communication technologies for interacting is increasing. An extension of these is mobile robotic telepresence (MRP) systems, video conferencing systems mounted on teleoperated mobile robots. The nature of the interaction via an MRP system is more complex than face-to-face interaction and involves not only social communication but also mobility. This research focuses on the use of MRP systems in domestic settings in elder care and contributes to the understanding of how interaction is affected by MRP system embodiment.

Does AI Need a New Debate on Ethics

Abstract: sometimes I watch TV in the evening. In September 2013, after the news, there was a preview of a talk show. The topic to be discussed was ‘The digital self—survival in the jungle of data’. Among the guests were the writer Hans Magnus Enzensberger and somebody I knew in person!—Stefan Wrobel. Stefan Wrobel is one of the most well-known researchers in German AI and he was formerly an editor of this journal. Of course, I stayed awake and watched the talk show. The buzz word of the show was Big Data. The biggest concerns discussed were, as to be expected, NSA and the security of personal data. Another concern was that everybody gets swamped with personalized advertisements. The most important problem, however, in my opinion, was raised by Hans Magnus Enzensberger: the danger of applying intelligent techniques to such data to predict personal conditions and behaviors such as financial power, state of health, or criminal conduct. He pointed out that people will only recognize the danger of this technology if they won’t get a job because their profile is classified into the wrong category. Other examples along this line are: being rejected by a health insurance company because medical history leads to a prediction of a severe and expensive illness in the future or getting arrested because of taking a walk in an area which is currently classified as potential meeting place for criminal elements. During my undergraduate days, there was some discussion of ethics of AI research in relation to military aims. Currently, a small group of AI-related researchers are discussing robot ethics in the context of technological singularity. Otherwise, most AI researchers, including myself, are focussed on making progress on intellectually fascinating and challenging problems such as autonomous learning, object recognition, context-awareness, or the special topic of this issue—transfer learning—and so on. Of course, AI researchers are trained to do AI research and not to think about the impact of AI technology on society. AI research can be proud that research has now matured so much that it can contribute to solve complex real world problems. Nevertheless, in my opinion, we should at least be aware of possible dangers in relation to certain areas of application, we should point these dangers out to our students—who might find jobs with insurance companies or banking institutes after their graduation—, and we should invite ethicians to a joint debate on the ethics of exploiting big data with AI technology. I would be very happy for contributions along this line in the discussion column of this journal.