Thursday Morning Talk

What can AI systems do? Answering this question requires us to model their capabilities, but this first demands a clear conception of what capabilities are and which tools we can use to measure them. I advance a dispositional account of capabilities, understanding them as a system’s propensity to behave in certain ways under certain conditions.

Far-reaching perspectives in swarm robotics consider robots that are minimalist in their sensing, communication and computation, but are deployed in thousands to collaborate towards the accomplishment of tasks distributed in space and time. Generally speaking, future robot swarms might face harsh operating conditions where little communication is possible and no external infrastructure is available. These

Science often deals with issues pertaining to possibilities, contingencies and necessities, by engaging in thought experiments and modeling. This talk discusses how much deep learning can be helpful for navigating the possibility space for intelligence, adding to our scientific understanding of possibilities. One epistemically useful feature of neural networks is their runnability - they can

More details to follow. Photo created with DALL-E by Maria Ott.

The human brain extracts complex information from visual inputs, including objects, their spatial and semantic interrelations, and their interactions with the environment. However, a quantitative approach to capture this information remains elusive. I will present work where we show that LLM embeddings of scene captions successfully characterise brain activity evoked by viewing the natural scenes.

More details to follow.   Photo by lance Anderson on Unsplash.

The study of animal behavior is rapidly changing due to recent advances in long-term recording and automated analysis. Here we use these new developments to characterize mouse behavior via their temporal statistics. We analyzed positional data (RFID detections) of groups of mice housed in complex environments over many months. We found that behavior spanning seconds

More details to follow. Photo by David Clode on Unsplash.

More details to follow. Image created with DALL-E by Maria Ott.

Abstract Minimalistic robot swarms hold great promise for applications in healthcare, disaster response, and environmental monitoring. A key challenge lies in enabling these robots to rapidly and reliably reach consensus using limited communication, computation, and memory. In this talk, we explore how robot swarms can collectively identify the best among multiple discrete options in their

Abstract: Cognitive neuroscience seeks theories that jointly explain behavioral, neural, and mental states. The dominant approach is to use specialized tasks designed to optimally probe a concept of interest (e.g., episodic memory), and to disentangle behavioral, sensory, and mnemonic factors through design (e.g., by constraining gaze during image recognition). I will present an alternative framework

With the rise of modern deep learning, neural networks have become an essential part of virtually every artificial intelligence system, making it difficult to imagine different models for intelligent behavior. In contrast, nature provides us with many different mechanisms for intelligent behavior, most of which we have yet to utilize. One such underinvestigated aspect of