This talk will explore essential considerations in designing autonomous robots that align with human values and intentions. A primary focus will be on the alignment challenge: ensuring that robots not only act in accordance with human goals and ethical standards but that have an internal structure that allows them to also adapt to shifting expectations within complex, real-world environments. The discussion will concentrate on the two simultaneous functions the robot needs to constantly perform, aligning its internal purposes to those of humans and grounding these purposes, and the downstream domain dependent goals that arise from them, in its own sensory motor system. The integration of large language models (LLMs) will be examined as a tool for improving the robot's ability to align with and understand human intentions in a more dynamic, nuanced manner, as well as a resource to introduce cultural knowledge that can enhance robotic efficiency. The focus will be on how LLMs can enable robots to incorporate cultural insights, ensuring that they not only perform tasks effectively but also engage in contextually aware communication and decision-making.

A conceptual and operational perspective is presented on how the notion of purpose can be used to align open-ended learning (OEL) agents with user-defined objectives. Starting from a theoretical framing of OEL (understood as a process of autonomous, incremental skill acquisition in open and evolving environments), the integration of external guidance is explored, with attention to preserving the agent’s autonomy. Purpose is introduced as a high-level, domain-independent directive that can bias learning dynamics toward goals more likely to fulfill user needs. This approach enables agents to remain open-ended in their exploration while progressively developing competencies that are both self-organized and user-relevant. The framework is illustrated through concrete examples, showing how purpose-driven mechanisms can be implemented in practice to support alignment across diverse and dynamic scenarios.

Talk abstract: TBA

As autonomous robots move beyond structured factory floors into open, human environments, they face an autonomy–alignment problem: how to preserve open‑ended learning (OEL) while ensuring that what is learned serves human practical purposes and values. This talk first presents a unifying purpose framework that treats purpose as a design primitive encoding what humans want a robot to learn, do, or avoid, independently of task domains. The framework formalises alignment, establishes necessary and sufficient conditions under which it holds, and decomposes the autonomy–alignment challenge into four tractable sub‑problems: aligning robot and human purposes, arbitrating among multiple purposes, grounding abstract purposes into domain‑specific goals, and acquiring the competencies to achieve them. The second half of the talk instantiates one pathway enabled by the framework: intrinsic purpose. I introduce Purpose‑Directed Open‑Ended Learning (POEL), which operationalises intrinsic purpose by speech‑based purpose, large‑language‑model reasoning, and computer vision to identify purpose-relevant objects in the scene. These estimates are then used to modulate intrinsic rewards and induce spatial exploration biases that steer the robot exploration toward user‑relevant interactions while retaining the breadth of OEL. In two simulated manipulation domains, POEL accelerates learning and achieves higher success on previously unseen, purpose‑aligned tasks than state‑of‑the‑art OEL baselines. Together, the framework and model demonstrate how ‘purpose in the loop’ can reconcile autonomy with alignment, offering a principled path to open‑ended robots that learn autonomously yet remain reliably useful and safe for people.