Author: Philipp Wicke

Content

Beyond the normal use of your senses, what else is possible? There have been different approaches to augment your senses. Here you will learn and experience more about space perception. How do you perceive space? What is space perception, and can it be changed? People’s experiences and skills vary greatly on how easily they can orient themselves in unknown environments. What happens if you always knew where north is, as well? You can try it with the feelSpace belt, which always points towards north via vibration. You will perform your own little experiments to aim for a new perception of space. 

Throughout the course and the whole IK it will be possible to lend out compass belts. 

Spacial perception can also be augmented through sound, e.g. as known from bats through echolocation. Also humans are capable of using echoes to learn a lot about their environment which has proven to be a helpful orientation skill for many blind people. But how do blind people learn to use echolocation? And can sighted people learn to use echolocation? We will present you the basics of how it is instructed. You will learn that you can hear more than you might guess.

1.     Session: Introduction

You will learn what sensory substitution and augmentation is and what is currently done to enable the use of information via unusual paths. We will present some examples and introduce to you the compass belt from feelSpace and the research behind it that proves how using this tool can help to gain a new sense of space. We will try out the compass belt and do some small experiments to experience the new “north-sense”.

2.     Session: From research to startup: feelSpace

You will learn more about the story of feelSpace, the startup that develops tactile belts to make navigation easy for everyone. In groups you will develop your own ideas on using the presented technology and plan little experiments that you conduct in this session and throughout the course. 

3.     Session: Echolocation

Some blind people use echolocation to “see” their environment. How do they do it and what is possible with this method? In this session we teach you the basics and we will do some practical exercises to show you how you can “sense” the space around you. This session can be visited independently to the others. 

4.     Session: Finalizing and presenting your sensory augmentation experiments

You have time to rap up your results and short presentations and present them to the group. We will discuss your ideas and approaches and finish with a feedback round to discuss the experiences of the course participants with the compass belt and the entrepreneurial approach.

Lecturer

Julia Wache studied Cognitive Science in Vienna and Potsdam. She finished her PhD in Trento working on the Emotion Recognition via physiological signals and mental effort in the context of using a tactile belt for orientation. In parallel she participated in the EIT Digital doctoral program to learn entrepreneurial skills. In 2016 she joined the feelSpace GmbH that develops and sells naviBelts, tactile navigation devices especially designed for the visually impaired.

Affiliation: feelSpace
Homepage: feelspace.de

Lio Franz studied Linguistics and Computer Science at the University of Bielefeld and graduated M.A. in Linguistics, while also teaching new students the ropes of their field. During the pandemic they worked at the health office, learning much about the processes and bureaucracy around health administration and surveillance. Since 2023 they are part of the feelSpace GmbH team, supporting their endeavor to enable a greater accessibility via sensory augmentation through the naviBelt; which can be used as training for a more accurate sense of space, and guide our customers as a navigation tool of independence.

Affiliation: feelSpace
Homepage: feelspace.de

Lecturer: Ivana Kajic
Fields: Artificial Intelligence, Machine Learning, Cognitive Science

Content

Recent years have seen a proliferation of machine learning models that are capable of producing high-quality images that faithfully depict concepts described using natural language. Such models can generate images that represent arbitrary objects, object attributes, and complex relations between objects. In this talk, I will show that despite these impressive advancements, such models can still struggle with relatively simple tasks. Specifically, I will demonstrate that even the most advanced models have only a rudimentary notion of number sense. Their ability to correctly generate a number of objects in an image is limited to small numbers, and it is highly dependent on the linguistic context the number term appears in. I will further highlight challenges associated with evaluation of different model capabilities, including evaluation of numerical reasoning, and talk about different automated approaches that can be used to evaluate models in a more interpretable way by leveraging existing tools in machine learning and cognitive science.

Literature

• Kajić, I., Wiles, O., Albuquerque, I., Bauer, M., Wang, S., Pont-Tuset, J., & Nematzadeh, A. (2024). Evaluating Numerical Reasoning in Text-to-Image Models. 38th Conference on Neural Information Processing Systems.
• Testolin, A., Hou, K., & Zorzi, M. (2024). Large-scale Generative AI Models Lack Visual Number Sense. arXiv preprint arXiv:2402.03328.
• Wiles, O., Zhang, C., Albuquerque, I., Kajić, I., Wang, S., Bugliarello, E., Onoe, Y., Knutsen, C., Rashtchian, C., Pont-Tuset, J. & Nematzadeh, A. (2024). Revisiting Text-to-Image Evaluation with Gecko: On Metrics, Prompts, and Human Ratings. arXiv preprint arXiv:2404.16820.
• Kajić, I., & Nematzadeh, A. (2023). Evaluating Visual Number Discrimination in Deep Neural Networks. In Proceedings of the Annual Meeting of the Cognitive Science Society (Vol. 45, No. 45).
• Nieder, A. (2020). The adaptive value of numerical competence. Trends in Ecology & Evolution, 35(7), 605-617.
• Dhariwal, P., & Nichol, A. (2021). Diffusion models beat GANs on image synthesis. Advances in neural information processing systems, 34, 8780-8794.

Lecturer

Ivana Kajić is a Senior Research Scientist at Google DeepMind in Montréal, Canada. Her research interests include applying methods and techniques from cognitive science in analysis and characterization of behavior of machine learning models. Specifically, this includes designing evaluation protocols, benchmarks and metrics to comprehensively understand capabilities and limitations of large vision-language models that in recent years have demonstrated strong performance in a variety of tasks. She completed her PhD thesis titled “Computational Mechanisms of Language Understanding and Use in the Brain and Behaviour” in 2020 at the University of Waterloo in Canada.

Affiliation: Google DeepMind
Homepage: www.ivanakajic.me

Lecturer: Catherine Sibert
Fields: Cognitive Modeling, Neuroscience, Cognitive Architectures

Content

The human brain is an incredibly complex system, and many approaches and perspectives are taken in an effort to better understand how it works. While this diversity of approaches has led to many insights, too often they are taken in isolation from one another, and it can be difficult to incorporate the findings from one perspective into others. Modeling-based approaches to understanding cognition, with their origins in computer science, artificial intelligence, and information processing, struggle to connect with researchers with a stronger neuroscience focus, as the models often do not incorporate or function at a higher level of abstraction than what is known about the biological brain. However, what we do understand about the brain is often at a lower level of abstraction than can be easily connected to the higher level cognitive components of models. In this course, we will discuss the strengths and weaknesses of the modeling approach, what kind of information we can actually extract from the brain, and some examples of how the two approaches can be combined to form a broader perspective of how to study the brain.

Lecturer

Catherine Sibert is an assistant professor of Human Computer Collaboration at the Faculty of Science and Engineering of the University of Groningen. Her research focuses on the use of AI tools and frameworks in the analysis of human brain data, with an emphasis on how models of whole-brain cognition can provide insight into the underlying mechanisms of thought. She is also interested in how understanding the brain can inform and improve AI systems to collaborate more effectively with human users.

Affiliation: University of Groningen
Homepage: www.catherinesibert.com

Lecturer: Logan Bentley, Harriet Crisp
Fields: Education, Artificial Intelligence, Machine Learning

Content

Join us for an interactive, hands-on workshop where you’ll design AI-powered interventions to tackle real-world challenges in early childhood education. In this course we will merge academic expertise with industry-driven insights – equipping and inspiring you to create solutions addressing the global learning crisis.

The course will begin with foundational instruction in core concepts and practices of early childhood education, emphasizing its critical importance in shaping equitable futures. Then we’ll explore the transformative potential of AI in education, presenting case studies from EIDU (https://www.eidu.com/) where traditional pedagogical interventions have been augmented with AI. These will include:
– Structured Pedagogy: How we leverage large language models (LLMs) to generate unique and relevant lesson plans that empower teachers.
– Personalized Learning: The use of machine learning to tailor individual students’ learning paths and maximize their learning outcomes.

With this foundation, we’ll move into an immersive design workshop where you prototype practical, AI-driven interventions targeting improvement for early learners’ literacy and numeracy. You’ll be guided with instruction on best practices, proven frameworks, and lessons from our own experience.

The learning crisis is a global challenge; solving it demands collaboration across disciplines, industries, and borders. Let’s explore together how we all can make a lasting impact.

Literature

• Friedberg, A (2023). Can A/B Testing at Scale Accelerate Learning Outcomes in Low- and Middle-Income Environments? (https://link.springer.com/chapter/10.1007/978-3-031-36336-8_119)
• Sun, C, Major, L, Moustafa, N, Daltry, R, and Freidberg, A (2024) Learner Agency in Personalised Content Recommendation: Investigating Its Impact in Kenyan Pre-primary Education. (https://link.springer.com/chapter/10.1007/978-3-031-64312-5_25)
• Sun, C, Major, L, Moustafa, N, Daltry, R, and Freidberg, A (2024) Teacher-AI Collaboration in Content Recommendation for Digital Personalised Learning among Pre-primary Learners in Kenya (https://dl.acm.org/doi/10.1145/3657604.3664662)

Lecturer

Logan is an accomplished software engineer with over a decade of experience in the education and gaming industries. He is an engineering manager at EIDU, where for the last years his team has been focused on improving learning outcomes for EIDU\’s EdTech interventions in low-income countries.

Affiliation: EIDU GmbH
Homepage: https://www.eidu.com/

Harriet received her Master’s degree in Engineering from the University of Cambridge. She began her career as a Machine Learning Engineer at a data consultancy, where she worked for two years. She then joined EIDU as a Data Scientist, bringing her expertise to the field of educational technology.

Affiliation: EIDU GmbH
Homepage: https://www.eidu.com/

Lecturer: Marieke van Vugt
Fields: Psychology, neuroscience, cognitive science

Content

Contemplative practices such as mindfulness are often marketed as methods to augment our minds: making us more concentrated, more happy, more efficient. But is that all these practices are about? And what is actually the evidence? In contemplative science we critically examine how contemplative practices affect mind, brain and body. In this course, we will blend contemplative science with actual practice of the relevant contemplative practice. It seeks to give a broad overview of different practices. In session 1, we will practice mindfulness and discuss evidence for effect of mindfulness on cognition. In session 2, we will practice analytical meditation and discuss how it affects mind and brain. In session 3, we will focus on embodied practices and use contemplative dance as an example. There is relatively little science on this topic, but we will focus on the science of embodied practices in general, and leave space to discuss your own ideas about how to study these practices and/or how to use them in your own life.

Literature

• Fox, K. C., Dixon, M. L., Nijeboer, S., Girn, M., Floman, J. L., Lifshitz, M., … & Christoff, K. (2016). Functional neuroanatomy of meditation: A review and meta-analysis of 78 functional neuroimaging investigations. Neuroscience & Biobehavioral Reviews, 65, 208-228.
• Vago, D. R., & Silbersweig, D. A. (2012). Self-awareness, self-regulation, and self-transcendence (S-ART): a framework for understanding the neurobiological mechanisms of mindfulness. Frontiers in human neuroscience, 6, 296.

Lecturer

Marieke van Vugt received her PhD from the University of Pennsylvania. She is now an associate professor at the AI department of the University of Groningen, the Netherlands. In her lab, she tries to understand when, how and why we mind-wander, using methods from psychology, neuroscience and AI. She is also interested in the effects of contemplative practices on our mind, especially on our mind-wandering. In addition, she collaborates with Tibetan Buddhist monks on the practice of analytical meditation and monastic debate. Besides her work as an academic, she is also a classical ballet dancer with Amsterdam Amateur Ballet.

Affiliation: University of Groningen
Homepage: https://mkvanvugt.wordpress.com

Lecturer: Cosima Prahm
Fields: Neuroscience, Medicine, Engineering

Content

Although the hand represents only 1% of our body weight, most of our sensorimotor cortex is associated with its control. The loss of a hand therefore not only signifies the loss of the most important tool with which we can interact with our environment, but also leaves us with a drastic sensory-motor deficit that challenges our central nervous system. Restoring hand function is therefore not only an essential part of restoring physical integrity and functional employability, but also closes the neural circuit, thereby reducing phantom sensations and nerve pain.

When there is no longer sufficient anatomy to restore meaningful function, we can resort to complex robotic replacements whose functional capabilities in some respects even surpass biological alternatives, such as conservative reconstructive measures or transplantation of a hand. However, as with replantation and transplantation, the challenge with bionic robotic replacements is to solidly attach it to the skeleton and connect the prosthesis to our neural and muscular system to achieve natural, intuitive control and also provide basic sensory feedback.

This interdisciplinary course will discuss the progressive development of upper extremity robotic prosthetics in the fields of bioengineering, medicine, computer science, and neuroscience. We address the medical basis of biosignals, movement, amputation and restoration, and various systems of prosthetic limbs to restore physical integrity. We will discuss enhancement versus restoration and how to improve the man-machine-interface, exemplified with case studies. 

Sessions in detail: We will cover the fundamentals of bionic extremity reconstruction, including replantation, transplantation, and biological restoration, along with an overview of various upper limb prosthetic types. Participants will explore conventional and advanced prosthetic control methods, including Machine Learning applications and innovative human-machine interfaces like nerve transfers, osseointegration, and implanted electrodes. Cutting-edge concepts, such as self-contained neuromusculoskeletal prostheses and cyborg technologies, will also be discussed. The course will address virtual rehabilitation using mixed reality (XR) environments, focusing on traumatic hand injuries, phantom limb pain, and neural pain management. 

Literature

• Aszmann, O. C., & Farina, D. (2021). Bionic Limb Reconstruction. In O. C. Aszmann & D. Farina (Eds.), Bionic Limb Reconstruction (1st ed.). Springer International Publishing. https://doi.org/10.1007/978-3-030-60746-3
• Prahm, C., Daigeler, A., & Kolbenschlag, J. (2021). Bionische Rekonstruktion der oberen Extremität. In Plastische Chirurgie (3rd ed., pp. 135–145). Kaden.

Lecturer

Cosima Prahm received her PhD in Medicine – Clinical Neuroscience from the Clinical Laboratory for Bionic Extremity Reconstruction at the Medical University of Vienna, Austria. From 2019 to 2024, she led the Research Laboratory in the Department of Hand, Plastic, Reconstructive, and Burn Surgery at the University Clinic of Tübingen, BG Klinik, Germany. Currently, she serves as the Director of the Center for Clinical Research at the ukb Unfallkrankenhaus Berlin, within the Department of Hand, Replantation, and Microsurgery, Charité University Medicine, Berlin. Her research focuses on enhancing human-machine interfaces for upper extremity amputees, advanced prosthetic control, bionic reconstruction, nerve regeneration, organ-on-chip models, and virtual rehabilitation in XR environments.  

Affiliation: 
– Director of the Center for Clinical Research
– Clinic for Hand, Replantation and Microsurgery
– ukb Klinik Trauma Hospital Berlin, Germany
– Charité University Medicine Berlin, Germany

Homepage: www.playbionic.org

Lecturer: Magda Gregorova
Fields: Machine Learning, Deep Learning

Content

In this introductory course we will cover the basics of machine learning (ML) targeting specifically the un-initiated audience. If you are not sure what machine learning actually is, if you have never trained an ML model, if for you deep learning means learning in deep sleep and ChatGPT is a result of dark magic, then this course is meant for you. The course will be organized in four sessions where (1) we will begin from the basic concepts of learning from data reviewing the fundamental ideas and building stones of machine learning, (2) we shall discuss some classical ML algorithms which are still the workhorses for solving many practical problems, (3) we shall explore the more modern deep learning approaches based on neural network models, and (4) we shall uncover some of the magic behind ChatGPT by discussing the concepts of deep generative modelling. The field is vast and very fast paced combining mathematics with computer science while spicing it up with ideas coming from physics, neuroscience, and many other areas. While some math cannot be avoided, it is not my ambition to cover all the technicalities of ML. I will rather endeavor to help you build your own picture of the field based on basic understanding of the underlying fundamental ideas and nature your own intuition for data analysis hoping you will become more comfortable and confident when exploring the ML methods in your future work.

Literature

• Bishop, C. M. (2006). Pattern recognition and machine learning. Springer.
• Murphy, K. P. (2012). Machine learning: A probabilistic perspective. MIT Press.
• Hastie, T., Tibshirani, R., & Friedman, J. (2001). The Elements of Statistical Learning. Springer New York Inc.
• Shalev-Shwartz, S., & Ben-David, S. (2014). Understanding Machine Learning: From Theory to Algorithms (1st ed.). Cambridge University Press. https://doi.org/10.1017/CBO9781107298019
• MacKay, D. J. C. (2003). Information theory, inference and learning algorithms. Cambridge University Press.
• Cover, T.M. and Thomas, J.A. (2006) Elements of Information Theory. John Wiley & Sons, Inc., Hoboken.
• Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep Learning. MIT Press.
• Zhang, A., Lipton, Z. C., Li, M., & Smola, A. J. (2021). Dive into Deep Learning. ArXiv Preprint ArXiv:2106.11342.
• Prince, S. J. D. (2023). Understanding deep learning. The MIT Press. http://udlbook.com
• Tomczak, J. M. (2022). Deep Generative Modeling. Springer International Publishing. https://doi.org/10.1007/978-3-030-93158-2

Lecturer

Magda Gregorova comes from Prague, Czech Republic, where she obtained her Master‘s degree in Statistics (2001) from the University of Economics. She started her career as an applied statistician in the Czech National Bank, where she headed a technical unit on financial statistics and collaborated closely with the ECB and the IMF. After several years in banking she has decided to follow an international career and joined Eurocontrol, the European Organization for the Safety of Air Navigation based in Brussels, Belgium, as a statistical analyst and forecaster. She then moved to Geneva, Switzerland, where she obtained in 2018 a PhD in machine learning from the Computer Science Department of the University of Geneva. She continued as a post-doc in the Data Mining and Machine Learning group of the University of Applied Sciences of Western Switzerland. In 2021 Magda has moved to Germany, where she obtained the research professorship for “Representation and Learning in Artificial Intelligence” at the Faculty of Computer Science and Business Information Systems of the Technical University of Applied Sciences Würzburg-Schweinfurt (THWS). She is a founding member of the THWS research Center for Artificial Intelligence (CAIRO) which she has led from its beginnings in 2022 till mid 2024. Her teaching activities are mainly within the international masters on AI in the areas of deep learning and generative modelling. In her research she focuses on deep unsupervised learning methods for modelling complex high-dimensional distributions and data representations for downstream tasks (https://scholar.google.com/citations?user=68MKCOwAAAAJ&hl=en). In addition to her own research she regularly contributes to the machine learning community through reviewing service (ICML, ICLR, NeurIPS, etc.) and by active participation in outreach and educational events such as IK.

Affiliation: Technische Hochschule Würzburg-Schweinfurt
Homepage: https://fiw.thws.de/en/our-faculty/staff/person/prof-dr-magda-gregorova/

Lecturer: Alessandro Del Vecchio
Fields: AI, Neuroscience, BCI

Content

Spinal motor neurons represent the final gateway from neural intention to physical movement, making them crucial for any interface that aims to restore or augment motor function. In cases of spinal cord injury (SCI), paralysis of the hand muscles significantly impacts quality of life, as individuals lose the ability to perform fundamental tasks. However, our recent research demonstrates that even in individuals with motor complete SCI (C5–C6), the activity of spinal motor neurons remains accessible and task-modulatable. Using a minimally-invasive electromygraphic interface, we tested eight SCI individuals and identified distinct groups of motor units under voluntary control that could encode specific hand movements, from grasping to individual finger flexion and extension. By mapping these motor unit discharges to a virtual hand interface, we enabled participants to proportionally control multiple degrees of freedom, successfully matching various cued hand postures. These findings underscore the potential of wearable muscle sensors to access voluntarily controlled motor neurons in SCI populations, presenting a pathway to restore lost motor functions through assistive technologies.

Alongside this study, we explored the neural organization of motor unit activity in different muscle groups, focusing on the low-dimensional latent structures—or motor unit modes—that underlie the coordinated output of motor units. By applying factor analysis, we identified two primary motor unit modes that captured most of the variability in motor unit discharge rates across knee extensor and hand muscles. Interestingly, we observed a distinct pattern in the hand muscles, where motor unit modes were largely specific to individual muscles, whereas knee extensors displayed a more continuous distribution, with shared synaptic inputs leading to overlapping motor unit modes across muscle groups. Simulations with large populations of integrate-and-fire neurons confirmed the accuracy of these modes, shedding light on the common inputs that drive correlated activity in synergistic muscle groups.

Building on these insights, we have now developed an open-source software platform that translates real-time EMG activity into controllable movement outputs. This software seamlessly integrates with both exoskeletons and prosthetics, allowing for precise and intuitive movement control that aligns with the user’s intent. With this tool, we can now bring intuitive neuromotor interfaces closer to clinical reality, offering individuals with SCI and other neuromuscular impairments a new level of interaction and independence.

Literature

• Oliveira et al. Brain 2024 https://academic.oup.com/brain/article/147/10/3583/7631714
• Del Vecchio et al. J Neurosci 2023 https://pubmed.ncbi.nlm.nih.gov/36922028/
• Simpetru et al. arXiv 2024 https://arxiv.org/abs/2408.07817

Lecturer

Prof. Del Vecchio leads the n-squared lab (neuromuscular physiology and neural interfacing) at FAU since 2020 in the Dpt of AI in Biomedical Engineeirng. He is mainly interested in motor unit physiology, neuromotor interfaces, and machine learning.

Affiliation: FAU Erlangen-Nurnberg
Homepage: https://www.nsquared.tf.fau.de/

Lecturer: Jutta Kretzberg
Fields: Personal development / career advice

Content

Are you a student? Are you thinking about a PhD? And maybe even about a career in academia?
Does “doing a PhD” sound like fun? Or rather like pain?
Many Master students struggle with the decision if a PhD would be the right choice for their career. And a considerable percentage of PhD students continue wondering if their decision was right until they graduate (or even beyond).
There is no general advice who should pursue a PhD. The decision for or against a PhD is a personal one – it depends on many factors including your personality, your personal situation and the available job opportunities. The goal of this workshop is to help you to develop a clearer personal perspective on this decision.

Session 1: External perspectives
In the first session, I will start with a brief overview of different options how to do a PhD in Germany. After that, we will interactively explore the perspectives of different stakeholders: What do Master students expect from doing a PhD? What do PhD supervisors expect from their PhD students? What do employers expect from applicants with a PhD versus a Master’s degree? And what is the perspective of family and friends?

Session 2: Your personal perspective
In preparation of the second session, you will write cards with your hopes, neutral expectations, and fears concerning yourself doing a PhD. Teaming up with one of the other participants, both of you will cluster these cards into the categories “tasks / skills”, “topics / scientific questions”, “working environment”, and “personal factors”. Explaining your thoughts (and listening to the thoughts of your teammate) can sharpen your personal perspective and help you to identify core aspects of your career decisions.

Please note: This workshop consists of two sessions and will be offered two times for a maximum of 20 participants each. Please register for one of the two workshop iterations via the list at ‘Glaskasten’ during IK.
The main target group of this workshop are Master’s and advanced Bachelor’s students. However, the method to develop your personal perspective is also applicable to further career steps. PhD students, PhD holders, and positive non-PhDs who are willing to share their perspectives are highly welcome!

Literature

• Chris Woolston: “Graduate survey: A love-hurt relationship” Nature 550, 549-552 (2017) https://www.nature.com/articles/nj7677-549a https://doi.org/10.1038/nj7677-549a (Nature’s survey of more than 5,700 doctoral students worldwide)
• Lars Kiewidt, PhD: “To PhD or not to PhD?” (2019) https://medium.com/age-of-awareness/to-phd-or-not-to-phd-4312cdb862c5 (Evaluation of this survey data set concerning PhD student’s motivation, skills and satisfaction across fields in natural sciences.)
• Chris Woolston: ‘I don’t want this kind of life’: graduate students question career options Nature 611, 413-416 (2022) doi: https://doi.org/10.1038/d41586-022-03586-8 (Newest version of nature’s PhD survey, but not open access)
• Katie Mitzelfelt, PhD: “To Be or Not To Be a PhD Candidate, That Is the Question” (Association for Women in Science Magazine, 2021): https://awis.org/to-phd-or-not-phd/ (Individual perspectives of 6 persons on their own decision to be or not to be a PhD.)
• Charlotte King_: “To PhD or not to PhD, that is the question…” https://www.postgrad.com/blog/to-phd-or-not-to-phd/ (Rather old, but still helpful blog post)

Lecturer

Jutta Kretzberg is professor for Computational Neuroscience and head of the MSc program Neuroscience at University of Oldenburg. She studied applied computer science and biology at University of Bielefeld, where she also did her PhD in Biology. After being a postdoc (and having a baby) in San Diego, California, she came back to Germany to be a junior professor and became a professor some years (and another baby) later. Nowadays, while juggling her family, teaching, research and administration duties, her favorite task is mentoring.

Affiliation: University of Oldenburg, Germany
Homepage: https://uol.de/en/neurosciences/compneuro

Lecturer: Constantin Rothkopf
Fields: Cognitive Science, Artificial Intelligence

Content

Recent advances in artificial intelligence based on deep learning have led to the discovery of new medical drugs, the development of new materials, and the optimization of fusion reactor designs. However, claims about fundamental limitations persist: unpredictable blunders, limited robustness, and lack of explainability. The talk will present recent examples and studies contributing to the current debate on what current AI systems can do and what they cannot yet do. A central topic will be how to leverage Cognitive Science to understand the properties of such AI systems. The systems discussed include large language models, neural network models of economic decision-making, visual-language foundation models and the considered tasks range from the classic Bongard problems to sensorimotor control and planning under uncertainty to deontological ethical judgments. Topics will cover the anthropomorphization of AI systems, problems of data contamination and bias, Clever-Hans phenomena, inherent limitations of benchmarks, and fundamental limitations of evaluations and comparisons in terms of performance measures of behavior.

Literature

• Schramowski, P., Turan, C., Andersen, N., Rothkopf, C. A., & Kersting, K. (2022). Large pre-trained language models contain human-like biases of what is right and wrong to do. Nature Machine Intelligence, 4(3), 258-268.
• Binz, M., & Schulz, E. (2023). Using cognitive psychology to understand GPT-3. Proceedings of the National Academy of Sciences, 120(6), e2218523120.
• Mitchell, M. (2023). How do we know how smart AI systems are? Science, 381(6654), eadj5957.
• Valmeekam, K., Marquez, M., Sreedharan, S., & Kambhampati, S. (2023). On the planning abilities of large language models-a critical investigation. Advances in Neural Information Processing Systems, 36, 75993-76005.
• Thomas, T., Straub, D., Tatai, F., Shene, M., Tosik, T., Kersting, K., & Rothkopf, C. A. (2024). Modelling dataset bias in machine-learned theories of economic decision-making. Nature Human Behaviour, 8(4), 679-691.
• McCoy, R. T., Yao, S., Friedman, D., Hardy, M. D., & Griffiths, T. L. (2024). Embers of autoregression show how large language models are shaped by the problem they are trained to solve. Proceedings of the National Academy of Sciences, 121(41), e2322420121.
• Wüst, A., Tobiasch, T., Helff, L. , Singh Dhami, D., Rothkopf, C. A., Kersting, K. (2024). Bongard in wonderland: visual puzzles that still make AI go mad? Sys2-Reasoning, Neurips Workshops.

Lecturer

Constantin Rothkopf is a full professor (W3) at the Institute of Psychology in the Department of Human Sciences with a secondary appointment in the Department of Computer Science at the Technical University of Darmstadt. He is the founding director of the Center for Cognitive Science and a founding member as well as a member of the executive board of the Hessian Center for Artificial Intelligence (hessian.AI). He is also a member of the board of directors of the Center for Mind, Brain and Behavior (CMBB). He is a member of the European Laboratory for Learning and Intelligent Systems (ELLIS), a faculty member of the ELLIS Unit Darmstadt, and a member of the DAAD Konrad Zuse Schools of Excellence in Artificial Intelligence (ELIZA). He is currently co-speaker of the collaborative projects The Adaptive Mind and Whitebox. After obtaining a joint PhD in Brain & Cognitive Sciences and Computer Science at the Center for Visual Science at the University of Rochester, NY in 2009, he started a postdoc at the Frankfurt Institute for Advanced Studies (FIAS) working in the theoretical neuroscience group. In 2009 he started as a lecturer at the Goethe University, Frankfurt, and from 2010 to 2012 he was the principal investigator of the “beliefs, representations, and actions group” at FIAS. After a year as a substitute professor at the Institute of Cognitive Science at the University Osnabrück, he started as an associate professor for “psychology of information processing” at the Institute of Psychology at the Technical University of Darmstadt in 2013. During the winter semester 2017 he was a visiting professor at the Department of Cognitive Science at the Central European University, Budapest. In 2022 he received an ERC Consolidator Grant from the European Research Council for his project ‘ACTOR’. During the summer semester 2023 he was a visiting professor at the Zuckerman Institute, Columbia University, New York, USA.

Affiliation: TU Darmstadt
Homepage: https://www.pip.tu-darmstadt.de