Author: Philipp Wicke

Lecturer: David Burden
Fields: Artificial Intelligence, Metaverse, Embodied Intelligence, Digital Immortality

Content

Stories and comment around digital immortality seem to be increasingly common in the media, and public reaction (particularly in the West) is often very negative. But what is the true current situation, and how might things evolve in the future? This series of lectures will explore the emergence and evolution of both virtual human and metaverse technology, and how they come together in the form of embodied AI. One possibility, almost an implication of these developments is the creation of digital replicas of real human beings, as well as of synthetic personalities, and how once such entities emerge they may deliberately, or even accidentally, transcend the death of their human models, and take on a form of digital amortality. The lectures will consider both the technical possibilities and the ethical and social implications of such developments. Students will be directed to resources which will enable them to experiment with the technologies discussed, to prompt discussions and explore the future through a matrix or similar narrative game.

Lecture 1 – Conversational AI and Virtual Humans

Lecture 1 will examine at the development of Conversational AI and virtual humans, from Eliza to ChatGPT. In particular it will look at the different elements required to create the ‘mind’ and ‘body’ of a virtual human, and how Large Language Model approaches contrast with some of the more traditional approaches, and whether hybrid approaches are possible or desirable. The lecture will also discuss the current use cases for conversational AI and the challenges, particularly in terms of ethics, privacy and employment. The representation of such AIs in the media, such as in Spike Jonze’s Her will also be considered. The idea of personal replicas will be introduced and attendees will be introduced to tools that will allow them to build their own personal replica during or after the School.

Lecture 2 – The Metaverse

Lecture 2 will explore the evolution of virtual worlds and metaversal spaces, with a particular focus on social virtual worlds. Current models of virtual worlds, such as Fortnite, Robolox etc will be compared to those of almost a generation ago such as There, ActiveWorlds and Second Life. Current use cases and challenges will be considered. Different definitions of the metaverse will be discussed and ten axioms for a future metaverse considered. The lecture will also examine 7 possible futures for the concept of the Metaverse. The idea of information gardens, a memory-palace or second brain made explicit, will also be explored. Media representation of the Metaverse will also be presented, such as Vernor Vinge’s True Names, Ernest Cline’s Ready Player One, Lottie Moggach’s Kiss Me First and the Mindjammer Role-Playing Game (RPG). Attendees will be introduced to tools that will allow them to build their own information gardens during or after the School.

Lecture 3 – Embodiment

Lecture 3 will bring together the two strands of ideas and technology presented in the previous lectures by considering the role of embodiment in the development of intelligence and sentience. The lecture will examine the embodiment of AI in physical (robot/android) and virtual (avatar/robotar) forms, how they compare and their relative affordances. The virtual embodiment model will then be considered in more detail, looking at past and present examples, and future directions for the research, and even dreaming (of electric sheep?) will be considered. Media representation of embodiment, from the physical embodiment of Blade Runner and Battlestar Galactica’s Caprica prequel and the Sentience RPG to Andromeda’s Rommie and the fluid embodiment of Greg Egan’s books will be examined. Attendees will (hopefully) be introduced to tools that will allow them to create their own embodied robotars during or after the School.

Lecture 4 – Digital Amortality

Lecture 4 will push on into the future and, informed by the previous lectures, consider the question of digital amortality – is it viable (or even inevitable) and is it desirable. Digital amortality (a term coined by in Neal Stephenson in Fall; or, Dodge in Hell) is a slightly more ‘realistic’ version of digital immortality where it is accepted that true immortality may be something of a tall order, and that a more reasonable goal may be to have a control over when to end (or pause) one’s existence. Whilst a possible technical pathway will be presented, much of this session will be about the ethical and motivational issues for such developments, and particularly the different perspectives (and goals) of the person themselves and of those left behind. An argument as to why such amortality may be accidental, and inevitable will be presented, along with a consideration of the concept of a thanoverse, and the relationship between digital amortality and space exploration. There will be some live polling to gain attendee feedback on the different issues, and options, discussed, media representations (such as Black Mirror and Upload) will again be examined, and a potential personal pathway towards digital amortality presented.

Evening Session – A Matrix Game of the Future (provisional)

As an evening session a ~2 hour “matrix” game will be presented, focussed on the future development of AI, the metaverse and digital immortality. A Matrix Game is a form of structured argument. Each player takes on the role of a different stakeholder (e.g. country, organisation, pressure group, AI?) in the topic under discussion, and each turn presents an argument for undertaking one action – which can be challenged or supported by other players. A suitable adjudication method (e.g. dice, voting, probability cards) is then used to decide if the action is successful. For this game a turn may represent around 10-20 years so that the next century of human history, and of AI, metaverse and digital immortality can be examined.

Literature

• Burden, D. J., Savin-Baden, M., & Bhakta, R. (2016). Covert Implementations of the Turing Test: A More Level Playing Field? In Research and Development in Intelligent Systems XXXIII: Incorporating Applications and Innovations in Intelligent Systems: Vol. XXIV 33 (pp. 195–207). Springer International Publishing.
• Burden, D., & Savin-Baden, M. (2019). Virtual Humans: Today and Tomorrow. CRC Press.
• Savin-Baden, M., & Burden, D. (2019). Digital immortality and virtual humans. Postdigital Science and Education, 1, 87–103.
• Burden, D. J. H. (2020). Building a Digital Immortal. In M. Savin-Baden & V. Mason-Robbie (Eds.), Digital Afterlife. CRC.
• Burden, D., & Savin-Baden, M. (2024). The Metaverse: A Critical Introduction. Taylor & Francis.

Lecturer

Biography: For the last 20 years David has run Daden Limited, helping organisations explore and exploit the social and commercial potential of using conversational AI and virtual worlds, delivering over 100 projects for clients across the globe. Daden were finalists in the BCS Machine Intelligence Competition, and chatbots designed by David successfully passed two covert Turing Tests in the 2010s. David spoke at the inaugural TEDxBrum on Digital Immortality, has authored over a dozen papers and book chapters, including co-authoring the books Virtual Humans, and The Metaverse:A Critical Introduction, both published by Routledge, New York. David is currently studying for a PhD on wargaming urban conflict, is an ex-Royal Signals officer, a Chartered European Engineer and is also series co-editor for Taylor & Francis on their Metaverse Series of books.

Affiliation: Daden Limited
Homepage: https://taunoyen.com/

Lecturer: Benjamin Paassen
Fields: Machine Learning

Content

AI systems such as image generators, language models, automatic decision making systems, and much more are widely known. But what are the underlying models and algorithms that make these systems work? How does one take data as input and automatically extract models from them? This is the subject of machine learning.

The course will provide an introduction to machine learning. The core knowledge and skills taught by the course are:
– the basic recipe behind machine learning (training data, model architecture, loss function, training/optimization, and inference)
– the fundamental mathematical concepts behind machine learning
– example models and algorithms, from classic machine learning to neural networks
– types of machine learning (supervised, unsupervised, reinforcement)
– core notions for responsible machine learning, namely: interpretable models, adversarial examples, and fairness

In more detail, the course will have four sessions with the following topics:

1. Basic Concepts: Functions, learning algorithms, optimization, linear regression (as an example of a learning algorithm), regularization, probability theory, machine learning theory, how to design a ML experiment, how to read an ML paper
2. Recipes for interpretable and robust machine learning: Distance-based models, adversarial examples, and decision trees
3. Artificial neural networks and deep learning: Neural network modules, recipes for neural networks, generative models (diffusion and large language models)
4. Reinforcement learning and ethics

Each session is accompanied by a (voluntary) programming exercise in Python. Exercise sheets (and slides) can be found here: https://bpaassen.gitlab.io/Teaching.html

Literature

• Literature is optional and more regarded as ‘further/complementary reading’:
• Barber, D. (2012). Bayesian Reasoning and Machine Learning. Cambridge University Press. Cambridge, UK. http://www.cs.ucl.ac.uk/staff/d.barber/brml/
• Goodfellow, I., Bengio, Y., and Courville, A. (2016). Deep Learning. MIT Press. Cambridge, MA, USA. https://www.deeplearningbook.org/
• Paaßen, B., Artelt, A., Hammer, B. (2019). Lecture Notes on Applied Optimization. Bielefeld University. https://pub.uni-bielefeld.de/record/2935200
• Shalev-Shwartz, S., and Ben-David, S. (2014). Understanding Machine Learning: From Theory to Algorithms. Cambride University Press. Cambridge, UK. https://www.cs.huji.ac.il/w~shais/UnderstandingMachineLearning/
• Barocas, S., Hardt, M., and Narayanan, A. (2023). Fairness and Machine Learning. MIT Press. Cambridga, MA, USA. https://fairmlbook.org/

Lecturer

Benjamin Paaßen is Junior Professor for Knowledge Representation and Machine Learning at Bielefeld University and research affiliate at the Educational Technology Lab of the German Research Center for Artificial Intelligence (DFKI). Their research foci are interpretable machine learning, machine learning for education, and limitations of large language models (especially as research tools).

Affiliation: Bielefeld University
Homepage: https://bpaassen.gitlab.io/

Lecturer: Marieke van Vugt
Fields: Cognitive science/Contemplative science

Content

In this course, we will combine first- and third-person methods to explore mind-wandering, as well as meditation, which in some conceptions is a way to get to know one’s mind-wandering. We will learn about the scientific studies of mind-wandering and meditation, but also do some practice ourselves, and discussion about what we notice.
Session 1 will mainly explore mind-wandering, how mind-wandering is studied in the laboratory, and involve also a first-person observation of our own mind-wandering.
Session 2 will explore the content and phenomenology of mind-wandering, and how those determine its effects, for example in a psychiatric context. We will also try to change the content or phenomenology of our mind-wandering.
Session 3 will shift attention more to meditation, and we will discuss the different meditation practices that have been distinguished in the scientific literature, as well as trying them out. We will also bring some attention to the role of the body in mind-wandering.

Literature

• Smallwood, J., & Schooler, J. W. (2015). The science of mind wandering: Empirically navigating the stream of consciousness. Annual review of psychology, 66(1), 487-518.
• van Vugt, M. K., Soepa, J., Gyaltsen, J., Gyatso, K., Lodroe, T., Aadhentsang, T., … & Mishra, S. (2023). Using the body to think: an analysis of the cognitive mechanisms underlying Thinking at the Edge and Tibetan monastic debate.
• Kordeš, U., & Demšar, E. (2023). Horizons of becoming aware: Constructing a pragmatic-epistemological framework for empirical first-person research. Phenomenology and the cognitive sciences, 22(2), 339-367.

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: Ronald Sladky
Fields: Neuroscience, Cognitive Science, Modeling

Content

Originally, the idea of free energy minimization has been used as a tool for data modeling, in particular, to model effective connectivity in the brain based on neuroimaging data. On top of this, Karl Friston has proposed the free energy principle as a general principle for understanding brain functions. A bold statement – but, if true, one of the biggest breakthroughs in cognitive science.

In the form of active inference, the free energy principle could provide a neuro-computational explanation for predictive processing, the Bayesian brain hypothesis, and enactive cognition. It could provide a unified conceptual and computational framework to link previously distant and isolated research fields in the cognitive sciences that could be compatible with how we understand self-organization of living systems in a physical world.

In this course we will talk about how we make sense of data using models. By looking at how cognitive neuroscientists study the human brain using fMRI and brain connectivity methods. We discuss how models shape the way we interpret the world. I will teach you how these models work so you will find out on your own what to do with them. To what degree is the free energy principle useful for you? Modeling data, modeling the brain – or modeling the mind?

Session 1 covers how we turn fMRI data into brain activation and connectivity models. Session 2 focusses on dynamic causal modeling, an application of the Free Energy Principle, to study effective connectivity in the brain. Session 3 extends the Free Energy formalism used in DCM to describe brain functions and life as we know it. Session 4 will review state of the art applications, theoretical developments, and empirical evidence for the free energy principle and active inference in action. As an illustration, I will use my own research on amygdala functions and dysfunctions and its connectivity. So, we will also talk about fear, trust, and other emotions – not just data, brains, and methods.

Note: The course will try to prepare you to interface with Kathryn Nave\’s SC3 – Models and Metaphysics of Living Systems (https://interdisciplinary-college.org/2026-sc3/). This was a last-minute addition to the program. Some details are subject to ad hoc changes. I will probably skip parts of this very dense lecture and give you space for discussion and critical reflection.

Literature

• Parr, T., Pezzulo G., & Friston K. J. (2022). Active Inference: The Free Energy Principle in Mind, Brain, and Behavior. FREE e-book https://mitpress.mit.edu/9780262045353/active-inference/
• Friston, K. J. (2009). The free-energy principle: a rough guide to the brain?. Trends in cognitive sciences, 13(7), 293-301.
• Marreiros, A. C., Stephan, K. E., & Friston, K. J. (2010). Dynamic causal modeling. Scholarpedia, 5(7), 9568.
• Friston, K. (2013). Life as we know it. Journal of the Royal Society Interface, 10(86), 20130475.
• van Es, T. (2021). Living models or life modelled? On the use of models in the free energy principle. Adaptive Behavior, 29(3), 315-329.
• Corcoran, A. W., Hohwy, J., & Friston, K. J. (2023). Accelerating scientific progress through Bayesian adversarial collaboration. Neuron, 111(22), 3505-3516.
• Sladky, R., Kargl, D., Haubensak, W., & Lamm, C. (2024). An active inference perspective for the amygdala complex. Trends in Cognitive Sciences.

Lecturer

Ronald Sladky is a cognitive neuroscientist at the University of Vienna, Austria, at the Faculty of Psychology (Department of Cognition, Emotion, and Methods in Psychology). He teaches courses on cognitive science and predictive processing. His research focusses on the amygdala as well as emotion processing and social cognition in the human brain.

Affiliation: University of Vienna
Homepage: http://sweetneuron.at

Lecturer: Celeste Kidd
Fields: Cognitive Learning, Curiosity & Metacognition, AI & Misinformation

Content

In this lecture series, I will discuss our lab’s research about how people (and some non-human animals) come to know what they know about the world. The world is a sea of information too vast for anyone to acquire entirely. How do people navigate the information overload, and how do their decisions shape their knowledge and beliefs? We’ll discuss recent empirical work about the core cognitive systems that people use to guide their learning about the world—including attention, curiosity, and metacognition (thinking about thinking). We discuss the evidence that people play an active role in their own learning, starting in infancy and continuing through adulthood, and how many of these mechanisms are shared with non-human animals. We’ll talk about why we are curious about some things but not others, and how our past experiences and existing knowledge shape our future interests. We’ll also discuss why people sometimes hold beliefs that are inconsistent with evidence available in the world, and how we might leverage our knowledge of human curiosity and learning to design systems that better support access to truth and reality.

A running theme throughout this series will be the importance of uncertainty in guiding our learning and collaborative knowledge building. We will be taking a comparative approach in outlining the types of cognitive representations of uncertainty shared among biological intelligence, but lacking in artificial ones, to explain how current generative AI models cannot be trusted to disseminate information to people without problems. We’ll discuss several core tenets of human psychology that can help build a bridge of understanding about what is at stake when discussing regulation and policy options to prevent widespread adoption of these AI technologies from permanently distorting human beliefs in problematic ways.

Lecturer

Celeste Kidd studied print journalism and linguistics at the University of Southern California, where she earned a dual honors degree in 2007. Kidd moved to the University of Rochester for her graduate studies, where she worked in brain and cognitive studies and earned her PhD in 2013. She worked with Richard N. Aslin, an expert on infant learning. Kidd held visiting positions at Stanford University and the Massachusetts Institute of Technology. Kidd is a professor of psychology at the University of California, Berkeley.

Affiliation: UC Berkeley
Homepage: https://psychology.berkeley.edu/people/celeste-kidd

Lecturer: Terry Stewart
Fields: Computational Neuroscience

Modern AI deep-learning systems are very impressive, however they take a large amount of training data and can be very power-hungry. This large amount of training time can be thought of as an attempt to find good ways for the neural network to represent information. Biological brains, in contrast, tend to learn new tasks quickly, perhaps because they have already learned a good general-purpose neural representation of data. This means that if we can understand how brains represent information, we can then take those styles of representation and force AI systems to use exactly that same style of representation. In this talk I will show some examples of doing this, and show how it reduces training time and makes AI systems more robust.

Unfortunately, Lara Wilkin had to cancel due to sickness. Her course description is still available here:

ET1 – From a Creative User and Professional Peer Counselor to a Researcher in Upper Limb Prosthetics

Lecturer: Lara Wilkin
Fields: Research – Upper Limb Prosthetics, Professional Peer Counselor, Creative Background

Content

Growing beyond oneself as an expert in one\’s own field and looking beyond the familiar horizon is often initially overlooked until one\’s competence is proven. Research is interdisciplinary, and who, if not the users themselves, possesses the in-depth knowledge of everyday requirements and challenges?

By combining a natural-scientific and technical affinity with a creative background, Lara Wilkin is engaged both in research and through her voluntary work and networks to pave new ways, raise awareness, and advocate for those affected. Her journey spans from being a creative user and professional peer worker to becoming a researcher in the field of upper limb prosthetics.

Lecturer

Lara Wilkin – Academic and Professional Background Ms. Lara Wilkin possesses a comprehensive interdisciplinary expertise in the fields of upper limb prosthetics and design, which is complemented by her academic career, professional experience, and extensive further education and training. She is currently pursuing a PhD in Upper Limb Prosthetics at Bauhaus University (since 2024). Her professional career includes positions in various institutions and companies, including her work at Saphenus Medical Technology (2023–2025). Additionally, she completed advanced courses in rehabilitation technology, prosthetics, and biomedical engineering at TU Wien (2023–2024). Since 2021, Ms. Wilkin has been actively engaged as a volunteer prosthesis usage trainer for all hand prosthesis models and works as a workshop leader and speaker, addressing medical associations and prosthetic care professionals in collaboration with the Federal Ministry of Labour and Social Affairs, as well as presenting at the World Congress of OTWorld. In the field of advocacy and voluntary work, she assumed the role of Germany’s first official representative for upper limb amputations at the Federal Association for People with Arm or Leg Amputations (BMAB) in July 2022. From November 2023 to September 2024, she also served as Vice Chair of LVamp NRW e.V. Prior to that, she was a board member of the State Association for People with Arm or Leg Amputations NRW e.V. from February 2022 to November 2023. Ms. Wilkin is a functional bilateral upper limb amputee, having undergone a transradial amputation in 2019 and experiencing another accident in February 2022. Her personal experience allows her to integrate both scientific and technological perspectives into her work. Her dedication to prosthetics is reflected in the founding of the Arm Prosthesis Community (2021), where she serves as an expert, consultant, and network partner. Since 2020, she has worked as a research assistant in the field of information technology, where her activities deepened her interest in biomedical engineering, biomedical information technology, and robotics. Additionally, since 2020, she has been officially trained in peer counseling for amputees and has been active both nationally and internationally, advising patients, medical professionals, and relatives on amputation-related topics. Her academic journey began with a Bachelor of Arts (2014), followed by a Master of Arts (2020). Since 2014, she has also worked as a freelance communication designer, collaborating with esteemed clients such as FH Dortmund (including projects for NASA/JPL), the Federal Association of Digital Publishers and Newspaper Publishers e.V., Adobe, Schweppes, Nescafé, and other major companies. Her creative work has earned multiple international awards, with exhibitions and competition wins in Italy, the USA, and China. In recognition of her scientific and practical expertise, she has become a member of leading international and national professional associations, including: * ISPO – International Society for Prosthetics and Orthotics (Germany e.V.) * BMAB – Federal Association for People with Arm or Leg Amputations e.V. * LVamp NRW – State Association for People with Arm or Leg Amputations e.V. * Arm Prosthesis Community * World Design Consortium * IDC – International Design Club * IAD – International Association of Designers * ICCI – International Council of Creative Industries * ISPM – International Society of Product Manufacturers * IBSP – International Bureau of Service Providers * AIBA – Alliance of International Business Associations Through her continuous education, including participation in specialized congresses and training sessions, such as the 2nd Amputation Course at the University Hospital Hamburg for Surgeons, where osseointegration and various amputation techniques were tested on body donor, as well as international world and annual congresses (ISPO, BMAB, etc.), Ms. Wilkin has expanded her global network. Her interdisciplinary background, which combines creative expertise, hands-on technical knowledge, personal experience, and scientific proficiency, makes her one of the very few experts worldwide who can evaluate all commercially available hand prosthesis models both through personal testing as a user and from a scientific perspective.

Affiliation: Lara
Homepage: www.linkedin.com/in/larawilkin

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/