Abstract: Like most of our technologies today, AI dramatically increases the world's carbon footprint, thereby strengthening the severity of the coming downfall of life on the planet. In this talk, I propose that recent advances in large dimensional mathematics, and especially random matrices, could help AI engage in the future economic growth. This being said,

Abstract: Bees have a diverse instinctual repertoire that exceeds in complexity that of most vertebrates. This repertoire allows the social organisation of such feats as the construction of precisely hexagonal honeycombs, an exact climate control system inside their home, the provision of the hive with commodities that must be harvested over a large territory (nectar,

Abstract: Some of my past and current research looks at "decisions from  experience,” i.e., decisions based on the personally experienced outcomes of past choices, along the lines of reinforcement learning models and how such learning and updating is related to and differs from the way in which people and other intelligent agents use other sources of information,

Estimates from large-scale replication projects in psychology suggest that the majority of studies from top journals do not replicate. Using commonly accepted research methods, several academic fields amassed prolific, seemingly coherent literatures on phenomena that do not exist, such as extrasensory perception and depression candidate genes. Throughout the biomedical and life sciences, data detectives keep finding highly cited

In this presentation, I will report on the results of my work so far on the concept of intelligence, summarising some of the main points and proposals made, and opening the floor for open discussion about the topic.   The Zoom Link will be sent the day before the lecture.

Abstract: Self-supervised learning (SSL) consists in training neural network systems without using any human annotations. Typically, neural networks require large amounts of annotated data, which have limited their applications in fields where accessing these annotations is expensive or difficult. Moreover, manual annotations are biased towards a specific task and towards the annotator's own biases, which

Abstract Machine learning models have achieved stunning successes in the IID setting. Yet, beyond this setting, existing models still suffer from two grand challenges: brittle under covariate shift and inefficient for knowledge transfer. In this talk, I will introduce three approaches to tackle these challenges, namely self-supervised learning, causal representation learning, and test-time training. More

Abstract: What is perception? The most intuitive and influential answer to this question has long been the one given by David Marr: To see the world is “to know what is where by looking” - to transform light into representations of objects and their features, located somewhere ins pace. But is this all that perception