Science of Intelligence Fair 2023
The very first conference and exhibition of the Cluster of Excellence Science of Intelligence (SCIoI) in Berlin.
This conference is a unique opportunity for attendees to learn from leading experts, exchange ideas, and engage in live demonstrations through several exhibits straight from the labs! With its varied program, the conference reflects the interdisciplinary orientation of the work in the Cluster and covers a broad spectrum of contributions from different disciplines.
The event is aimed at people from the fields of science and politics who are interested in understanding the challenges of developing intelligent technologies.
Discover the underlying principles of intelligence, whether artificial or biological, and their application in creating intelligent technology in engaging keynotes and a panel discussion with renowned speakers, as well as in our exhibition.
During the event, scientists from Science of Intelligence will present their research in an exciting exhibition.
Meet the Swarm Robots
Taken on their own, the “kilobots” are simple creatures. They can only communicate over a distance of ten centimeters and process a small amount of data. They only develop their intelligence in a collective. Together, they solve relatively complex problems in “decentralized networks” without a central control unit – similar to a school of fish escaping a shark . SCIoI scientists analyze swarm behavior in nature and transfer the principles to algorithms that control mini robots. In experiments they observe how swarms of robots come to decisions – and how intelligence emerges in the collective. In future, this knowledge could contribute, for example, to optimizing autonomous driving.
Interact with a Fish Swarm
Drawing inspiration from research on social predators and their interactions with prey groups in schools of fish, this demonstrator aims to visualize the secrets of movement rules and decision-making in these fascinating creatures. As you walk through the projected fish swarm, witness how the collective responds to your movements. Just like a school of fish escaping a predator, the swarm adapts and navigates, displaying intelligence in their synchronized actions. The knowledge gained from this project is used to develop bio-inspired algorithms, enabling simulations and embodied robotics to better understand the principles governing biological systems and collective intelligence.
Kinematic Puzzle: Escape Room
Inspired by the intelligent problem-solving behavior observed in birds like cockatoos, this demonstrator brings together behavioral biology and robotic technology. Our mission is to understand the underlying principles that empower birds to master these complex tasks and translate that knowledge into our autonomous robotic system. This exhibit showcases the potential of robotics, where sensorimotor capabilities, computational principles, and internal representations converge to solve enigmatic challenges.
Inspired by the rules that shape fish and bird swarms, the project that underlies this demonstrator explores the complexities of social interactions. Robotic fish take the stage as social partners, replacing living fish in this setup. With adaptive real-time behavior, these “socially competent” robots lead and interact with their counterparts, showcasing their problem-solving skills. Our Reinforcement Learning environment trains robots to engage with our artificial fish models, unveiling strategies to tackle tasks effectively. The Robo fish holds immense research potential, fostering collaborations with biologists and roboticists. Join us to explore collective behavior and its connection to robotics and artificial intelligence.
This exhibit demonstrates, how our scientists use an innovative approach to extract meaningful motion information from complex animal behavior by combining traditional video and event-based camera systems. During this showcase, our scientists present the potential of event-based cameras, mimicking the human visual system’s transient pathway. Discover how this technology opens up new possibilities to study natural intelligence and unravel the wonders of animal behavior.
Socially Responsive Robots
Learn about how scientists from SCIoI explore the vital role of social responsiveness in knowledge transfer between humans and robots. Drawing inspiration from educational psychology, social responsiveness is recognized as a crucial aspect of successful knowledge transfer in human-human interactions. It involves accurately reading signals and appropriately reacting to them, enhancing communication and learning outcomes. Combining insights from computer vision and educational psychology, we identify key principles of socially responsive behaviors in humans. These principles serve as a foundation to design and develop socially responsive artificial systems. Find out more about the fascinating realm of human-robot interaction.
Human-like Face Perception
In humans, perceiving and categorizing faces plays a fundamental role in understanding others, influenced by a unique interplay of facial features and top-down influences from cognitive domains like expectations, memories, and contextual knowledge. However, synthetic systems often lack this human-like integration, relying solely on visual input. The project behind this demonstrator aims to bridge this gap, mirroring human face perception. Empirical observations, computational modeling, and advanced image analysis, provide information about how prior knowledge influences facial expression interpretation. These insights will empower humanoid robots to adapt to social perceptions and engage in natural face-to-face communication with humans.
Speakers and Panelists
Rasha Abdel Rahman is a Psychologist. Her research focuses on the neurocognitive bases of visual perception, semantic and social-emotional processing as well as verbal and non-verbal interactions. She is particularly interested in the multifaceted interplay of these core human faculties that enables us to understand our environment and to successfully navigate in a complex social world. Together with SCIoI colleagues from robotics and computer vision she investigates social communication between humans and artificial systems. In her research, Rasha employs behavioral and electrophysiological methods.
David Bierbach is a biologist working on topics that range from individual differences between conspecifics to large-scale collective behaviors. He integrates field-based studies with analytical and experimental approaches in the laboratory. Through his highly interdisciplinary work, David has developed several experimental techniques to study animal behavior in the most standardized ways, from video playbacks and computer animations to the use of robots. He investigates how fish use anticipation in their social interactions. The overall aim is to implement this knowledge to build better bio-mimetic robots and social interaction algorithms.
Palina Bartashevich is a computer mathematician and computer scientist who studies collective behavior in biological and artificial systems. Her research investigates predator-prey interactions, specifically focusing on group hunting fish in the wild. By developing computational models that are tested in simulations and can be embodied on robotic platforms, she aims to understand how predators manipulate and control their prey during the hunt. The insights from her models can enhance not only our understanding of biological systems but also contribute to the development of novel algorithms in robotics and artificial intelligence.
Manuel Baum researches how cockatoos and robots can explore, understand and solve complex mechanical puzzles. Robots and animals need to solve such tasks, but especially for mechanical problems the information required is often unknown until the agent actually starts to solve them. Which information is actually task-relevant, and how can a robot or animal gather that information? This is the subject of task-directed exploration and interactive perception. Together with colleagues from Berlin, Oxford and Vienna, Manuel researches these aspects of intelligence in the context of mechanical problem solving.
Oliver Brock, spokesperson of Science of Intelligence, represents the synthetic discipline robotics. He has extensive experience in building real-world robotic systems, contributing also to related disciplines, including perception and machine learning. Within the fields of robotics, he is a leader in leveraging collaborations with analytical disciplines, in particular psychology and behavioral biology, to work towards an understanding of embodied intelligence. His projects include topics such as lockbox problem-solving in animals and robots, soft-robotic hands, and the foundations required to enable robotic agents to autonomously perform complex tasks in dynamic environments.
Jonas Frenkel has a background in psychology and human factors. With a keen interest in the interplay between humans and machines, his research focuses on the field of human-robot interaction in general and social robotics in particular. Prior to joining SCIoI, he worked on developing robotic therapy scenarios for children with autism by using emotion-sensitive technology. At SCIoI, he develops computational models of nonverbal social behaviors in order to improve our understanding of the underlying principles of social interactions and to eventually allow synthetic agents to perceive and appropriately react to social cues. This understanding will in turn enable researchers to develop socially responsive artificial systems.
John-Dylan Haynes is an expert in several fields of cognitive neuroscience. He has worked on executive processes, such as the neural mechanisms of intentions and volitional control, on conscious and unconscious information processing and on information flow between brains during communication. He has also investigated neurotechnological applications, especially whether it is possible to decode thoughts from brain activity patterns. He is professor of theory and analysis of long-range brain signals at the Bernstein Center for Computational Neuroscience and at the Berlin Center for Advanced Neuroimaging (BCAN) of the Charité and HU Berlin and Principal Investigator at Science of Intelligence.
Verena Hafner is Professor of Adaptive Systems at Humboldt-Universität zu Berlin and Head of the Adaptive Systems Group at the Department of Computer Science. For Science of Intelligence, she represents the synthetic discipline of robotics and focuses on sensorimotor interaction and development. Verena has investigated open-ended development and social interaction in artificial agents that attracted high interest in the cognitive and developmental robotics community. Her work in robotics focused on developing intelligent algorithms and control strategies for autonomous robots, aiming to bridge the gap between perception and action in real-world environments.
Katharina Hohlbaum is a veterinarian with the focus on animal welfare, animal behavior, and laboratory animal science. Within SCIoI she investigates individual and social intelligence in mice, as well as these animals’ problem-solving strategies. Her data is used to develop synthetical simulations, which can raise new hypotheses regarding learning strategies of mice. Katharina is interested in the interaction of emotions and cognition. She aims to identify indicators of positive emotions and possible empathy in mice, as there is little known about their positive affective states. During her work, she developed a protocol on the systematic assessment of well-being in mice.
Alex Kacelnik FRS is a zoologist and behavioural ecologist who works on animal behavior and its underlying psychological mechanisms. His research includes studies of decision making, learning and memory in birds, mammals, insects, and other animals. With his work, he contributes to the integration of research strategies from analysis of animal behavior with the design of relatively autonomous artificial systems (robots). At Science of Intelligence, he collaborates with Oliver Brock and Alice Auersperg in their research on intelligence in cockatoos and its emulation in artificial systems. He is also is a member and promoter of the Oxford-Berlin academic partnership of the Berlin University Alliance.
Anna Lange possesses a diverse academic background that spans multiple fields, such as Mathematics, Psychology, and Computational Neuroscience. Her research is focused on human-robot Interactions using functional Magnetic Resonance Imaging (fMRI). Prior to SCIoI, she contributed to the development of neural networks inspired by human psychology for Robot-Robot Interactions. At SCIoI, Anna is working as a PhD for the Project 50. Her current research revolves around enhancing adaptive learning and teaching strategies in robots during social interactions. She achieves this by creating adaptive neural networks grounded in the principles of human psychology.
Martin Maier investigates face perception, bridging the gap between human and synthetic face processing by integrating top-down components typical for human perception into synthetic systems. Automatic face processing systems often rely solely on bottom-up information without incorporating prior knowledge, which differs from human perception and restricts the potential for understanding and effective interactions between artificial agents and humans. In his research, Martin creates paradigms to explore how humans and machines can draw on prior knowledge to improve visual perception, which, in perspective may be integrated in humanoid robots to adapt to social perception
Jörg Raisch’s research interests include both methodological and applied aspects of control. On the methodological side, he works on abstraction-based control synthesis for hybrid systems, control of timed discrete-event systems in tropical algebras, consistent control hierarchies, and distributed cooperative control. He has investigated control applications in chemical, medical and power systems engineering, with an emphasis on power systems with a high share of renewable energy. At SCIoI, he has worked on collective learning control in multi-robot systems, the trade-off between modularity and integration, and the use of regularities in complex control and decision problems.
Pawel Romanczuk studies self-organization, evolutionary adaptations, and functional dynamical behavior in animal collectives (such as flocks of birds or schooling fish,) and develops mathematical models to understand swarm behavior. He is concerned with determining how collective behavior is shaped and constrained by locally available sensory information, such as vision-based movement coordination as a response to the environment. Pawel explores how causal information flows can be identified within and between individual agents in animal groups, and how animal and human groups process this information. At Science of Intelligence, his work bridges analytical and synthetic sciences.
Christa Thöne-Reineke has extensive experience in laboratory animal science and animal models, especially in animal behavior as read out for severity assessment and animal welfare. She explores the costs and benefits of cognition and the influence of emotion and well-being on animal behavior. Christa’s research includes the analysis on animal models to gain an understanding of how social responsiveness is modulated by priors and emotions and how the model organism exhibits intelligent behavior. In perspective, this will allow computational modelling of subjective perception to be incorporated into synthetic agents such as robots. At SCIoI, Christa represents the analytical discipline of behavioral biology.