Search for: All records

We present an algorithm for skill discovery from expert demonstrations. The algorithm first utilizes Large Language Models (LLMs) to propose an initial segmentation of the trajectories. Following that, a hierarchical variational inference framework incorporates the LLM-generated segmentation information to discover reusable skills by merging trajectory segments. To further control the trade-off between compression and reusability, we introduce a novel auxiliary objective based on the Minimum Description Length principle that helps guide this skill discovery process. Our results demonstrate that agents equipped with our method are able to discover skills that help accelerate learning and outperform baseline skill learning approaches on new long-horizon tasks in BabyAI, a grid world navigation environment, as well as ALFRED, a household simulation environment. 

Bioelectronic medicine is emerging as a powerful approach for restoring lost endogenous functions and addressing life-altering maladies such as cardiac disorders. Systems that incorporate both modulation of cellular function and recording capabilities can enhance the utility of these approaches and their customization to the needs of each patient. Here we report an integrated optogenetic and bioelectronic platform for stable and long-term stimulation and monitoring of cardiomyocyte function in vitro. Optical inputs are achieved through the expression of a photoactivatable adenylyl cyclase, that when irradiated with blue light causes a dose-dependent and time-limited increase in the secondary messenger cyclic adenosine monophosphate with subsequent rise in autonomous cardiomyocyte beating rate. Bioelectronic readouts are obtained through a multi-electrode array that measures real-time electrophysiological responses at 32 spatially-distinct locations. Irradiation at 27 μW mm−2 results in a 14% elevation of the beating rate within 20–25 min, which remains stable for at least 2 h. The beating rate can be cycled through “on” and “off” light states, and its magnitude is a monotonic function of irradiation intensity. The integrated platform can be extended to stretchable and flexible substrates, and can open new avenues in bioelectronic medicine, including closed-loop systems for cardiac regulation and intervention, for example, in the context of arrythmias.