Shared autonomy combines human user and AI copilot actions to control complex systems such as robotic arms. When a task is challenging, requires high dimensional control, or is subject to corruption, shared autonomy can significantly increase task performance by using a trained copilot to effectively correct user actions in a manner consistent with the user’s goals. To significantly improve the performance of shared autonomy, we introduce Diffusion Sequence Copilots (DiSCo): a method of shared autonomy with diffusion policy that plans action sequences consistent with past user actions. DiSCo seeds and inpaints the diffusion process with user-provided actions with hyperparameters to balance conformity to expert actions, alignment with user intent, and perceived responsiveness. We demonstrate that DiSCo substantially improves task performance in simulated driving and robotic arm tasks. Project website: https://sites.google.com/view/disco-shared-autonomy/

Understanding and executing natural language instructions in a grounded domain is one of the hallmarks of artificial intelligence.In this paper, we focus on instruction understanding in the blocks world domain and investigate the language understanding abilities of two topperforming systems for the task.We aim to understand if the test performance of these models indicates an understanding of the spatial domain and of the natural language instructions relative to it, or whether they merely over-fit spurious signals in the dataset.We formulate a set of expectations one might have from an instruction following model and concretely characterize the different dimensions of robustness such a model should possess.Despite decent test performance, we find that state-of-the-art models fall short of these expectations and are extremely brittle.We then propose a learning strategy that involves data augmentation and show through extensive experiments that the proposed learning strategy yields models that are competitive on the original test set while satisfying our expectations much better.1 .

Wearable devices that capture users' rich information regarding their daily activities have unmet authentication needs. Today's solutions, which primarily rely on indirect authentication mechanisms via users' smartphones, thus cumbersome and susceptible to adversary intrusions. Even though there have been some efforts trying to fill this gap, they either rely on some superior sensors, such as cameras and electrocardiogram (ECG) pads, or are awkward to use, e.g., users are asked to perform some pre-defined movement/gesture for authentication. Therefore, an authentication mechanism for wearable devices that is accurate, robust, light-weight and convenient is in dire need.

The true power of genome editing mechanism known as RNA-programmable CRISPR Cas9 endonuclease system, lies in the fact that Cas9 can be guided to any loci complementary to a 20-nt RNA, single guide RNA (sgRNA), to cleave double stranded DNA, and therefore allows the introduction of wanted mutations. Unfortunately, sgRNA is prone to off-target homologous attachment, thus guiding Cas9 to cleave DNA sequences at unwanted sites. Using human genome and Streptococcus pyogenes Cas9 (SpCas9) as the example, this article analyzed the probabilities of off-target sites of sgRNAs and discovered that for large-size genomes such as human genome, off-target sites are nearly inevitable for sgRNA selection. Based on the mathematical analysis, it seems that the double nicking approach is currently the only feasible solution to promise genome editing specificity. An effective computational algorithm for off-target homology searching is also implemented to confirm the mathematical analysis.