Evolutionary & Hybrid AIat the VUB AI Lab
Our missionWe aim to build truly intelligent systems that are able to interact with and reason about their native environment in order to solve an open-ended set of tasks. Our systems are deeply inspired by evolutionary principles such as self-organisation, selection and emergent functionality, and are therefore adaptive by design. We adopt a hybrid approach that integrates symbolic and subsymbolic AI techniques, combining their strengths to achieve general, accurate and interpretable models. We focus in particular on tasks that require human language-like communication, involving advanced perception, reasoning and learning skills. We investigate fundamental research questions that have a tight connection to real-world problems.
Our expertiseOur expertise centres around three conceptual pillars that we believe to be crucial for achieving our goals. For each pillar, we design and develop computational tools and techniques that we deploy in our projects.
Hybrid engine for representing and processing symbolic and subsymbolic knowledge.
Meta-level learning for flexible knowledge expansion and robustly handling unforeseen situations.
Scalable multi-agent systems self-organising their concepts and language.
Featured projectsIn our projects we apply our in-house tools and techniques to achieve the best results for a given task.
Visual question answering
This project investigates how artificial agents can understand, reason about and answer natural language questions about images. We focus on the CLEVR task, adopting a hybrid and modular system architecture.
Origins of language
How can a population of agents self-organise a shared language that allows them to collaborate and solve a particular task? We study this question through agent-based models using physical robots in grounded scenarios.
Semantic frame extractor
We are building a semantic parser that is able to detect different ways in which events are framed in highly polarised debates, such as the climate change controversy. We retrieve FrameNet frames using our in-house hybrid frame-based processing engine.
SoftwareThe core software that we develop and apply across different projects is released under an open source license and hosted on GitHub.
Babel2A collection of software tools including a multi-agent simulation framework, a frame-based processing engine, metacognitive learning, procedural semantics engine and robot interfaces
PenelopeAn open modular software platform for the data-driven analysis and real-time tracking of opinion dynamics in online textual media aimed at social scientists.
Dr. Katrien BeulsSenior researcher and lecturer
Paul Van EeckePhD researcher
Jens NevensPhD researcher and teaching assistant
Interested in collaborating with us?
We're always looking for talented AI researchers to join our interdisciplinary teamGet in touch ›