The integration of autonomous vehicles (AVs) into transportation systems presents an unprecedented opportunity to enhance road safety and efficiency. However, understanding the interactions between AVs and human-driven vehicles (HVs) at intersections remains an open research question. This study aims to bridge this gap by examining behavioral differences and adaptations of AVs and HVs at unsignalized intersections by utilizing two comprehensive AV datasets from Waymo and Lyft. Using a systematic methodology, the research identifies and analyzes merging and crossing conflicts by calculating key safety and efficiency metrics, including time to collision (TTC), post-encroachment time (PET), maximum required deceleration (MRD), time advantage (TA), and speed and acceleration profiles. The findings reveal a paradox in mixed traffic flow: while AVs maintain larger safety margins, their conservative behavior can lead to unexpected situations for human drivers, potentially causing unsafe conditions. From a performance point of view, human drivers exhibit more consistent behavior when interacting with AVs versus other HVs, suggesting AVs may contribute to harmonizing traffic flow patterns. Moreover, notable differences were observed between Waymo and Lyft vehicles, which highlights the importance of considering manufacturer-specific AV behaviors in traffic modeling and management strategies for the safe integration of AVs. The processed dataset utilized in this study is openly published to foster the research on AV-HV interactions.

Despite the successes of large language models (LLMs), they exhibit significant drawbacks, particularly when processing long contexts. Their inference cost scales quadratically with respect to sequence length, making it expensive for deployment in some real-world text processing applications, such as retrieval-augmented generation (RAG). Additionally, LLMs also exhibit the "distraction phenomenon," where irrelevant context in the prompt degrades output quality. To address these drawbacks, we propose a novel RAG prompting methodology, superposition prompting, which can be directly applied to pre-trained transformer-based LLMs without the need for fine-tuning. At a high level, superposition prompting allows the LLM to process input documents in parallel prompt paths, discarding paths once they are deemed irrelevant. We demonstrate the capability of our method to simultaneously enhance time efficiency across a variety of question-answering benchmarks using multiple pre-trained LLMs. Furthermore, our technique significantly improves accuracy when the retrieved context is large relative the context the model was trained on. For example, our approach facilitates an 93x reduction in compute time while improving accuracy by 43\% on the NaturalQuestions-Open dataset with the MPT-7B instruction-tuned model over naive RAG.

Tactile sensors are believed to be essential in robotic manipulation, and prior works often rely on experts to reason the sensor feedback and design a controller. With the recent advancement in data-driven approaches, complicated manipulation can be realised, but an accurate and efficient tactile simulation is necessary for policy training. To this end, we present an approach to model a commonly used pressure sensor array in simulation and to train a tactile-based manipulation policy with sim-to-real transfer in mind. Each taxel in our model is represented as a mass-spring-damper system, in which the parameters are iteratively identified as plausible ranges. This allows a policy to be trained with domain randomisation which improves its robustness to different environments. Then, we introduce encoders to further align the critical tactile features in a latent space. Finally, our experiments answer questions on tactile-based manipulation, tactile modelling and sim-to-real performance.

GPT Summary: The rapid advancements in artificial intelligence (AI) have brought humanity to a critical juncture, similar to the Renaissance of the 14th-17th centuries. AI is experiencing a functional rebirth, with machines surpassing human performance in various cognitive tasks. These developments raise philosophical questions about the nature of human intelligence and our roles in a world where AI is omnipresent. Striking the right balance between innovation and regulation is crucial, as ethical concerns need addressing. By exploring AI's function and philosophical aspects, we can harness its power to enhance our lives, create new opportunities, and unlock the next renaissance in human-machine collaboration.

Robustly classifying ground infrastructure such as roads and street crossings is an essential task for mobile robots operating alongside pedestrians. While many semantic segmentation datasets are available for autonomous vehicles, models trained on such datasets exhibit a large domain gap when deployed on robots operating in pedestrian spaces. Manually annotating images recorded from pedestrian viewpoints is both expensive and time-consuming. To overcome this challenge, we propose TrackletMapper, a framework for annotating ground surface types such as sidewalks, roads, and street crossings from object tracklets without requiring human-annotated data. To this end, we project the robot ego-trajectory and the paths of other traffic participants into the ego-view camera images, creating sparse semantic annotations for multiple types of ground surfaces from which a ground segmentation model can be trained. We further show that the model can be self-distilled for additional performance benefits by aggregating a ground surface map and projecting it into the camera images, creating a denser set of training annotations compared to the sparse tracklet annotations. We qualitatively and quantitatively attest our findings on a novel large-scale dataset for mobile robots operating in pedestrian areas. Code and dataset will be made available at http://trackletmapper.cs.uni-freiburg.de.

In 1970, robotics expert Masahiro Mori first described the effect of the "uncanny valley," a concept that has had a massive impact on the field of robotics. The uncanny valley, or UV, effect, describes the positive and negative responses that human beings exhibit when they see human-like objects, specifically robots. The UV effect theorizes that our empathy towards a robot increases the more it looks and moves like a human. However, at some point, the robot or avatar becomes too lifelike, while still being unfamiliar. This confuses the brain's visual processing systems.

In task-oriented dialogue (ToD), a user holds a conversation with an artificial agent  with the aim of completing a concrete task. Although this technology represents one of  the central objectives of AI and has been the focus of ever more intense research and  development efforts, it is currently limited to a few narrow domains (e.g., food ordering,  ticket booking) and a handful of languages (e.g., English, Chinese). This work provides an  extensive overview of existing methods and resources in multilingual ToD as an entry point  to this exciting and emerging field. We find that the most critical factor preventing the  creation of truly multilingual ToD systems is the lack of datasets in most languages for  both training and evaluation. In fact, acquiring annotations or human feedback for each  component of modular systems or for data-hungry end-to-end systems is expensive and  tedious. Hence, state-of-the-art approaches to multilingual ToD mostly rely on (zero- or  few-shot) cross-lingual transfer from resource-rich languages (almost exclusively English),  either by means of (i) machine translation or (ii) multilingual representations. These  approaches are currently viable only for typologically similar languages and languages with  parallel / monolingual corpora available. On the other hand, their effectiveness beyond these  boundaries is doubtful or hard to assess due to the lack of linguistically diverse benchmarks  (especially for natural language generation and end-to-end evaluation). To overcome this  limitation, we draw parallels between components of the ToD pipeline and other NLP tasks,  which can inspire solutions for learning in low-resource scenarios. Finally, we list additional  challenges that multilinguality poses for related areas (such as speech, fluency in generated  text, and human-centred evaluation), and indicate future directions that hold promise to  further expand language coverage and dialogue capabilities of current ToD systems. 

"I love it when a plan comes together" – John (Hannibal) Smith, The A Team One of the biggest challenges that I continue to see are organizations struggling to cross the Analytics Chasm; to transition their use of data and analytics from retrospective Business Intelligence that monitors what has happened to AI/ML-driven analytics that predict what is likely to happen and prescribe preventative, corrective or monetization actions. I wrote about the challenges to crossing the Analytics Chasm in the blog "Crossing the Big Data / Data Science Analytics Chasm" and in further detail in my just released book "The Economics of Data, Analytics, and Digital Transformation". The Analytics Chasm prevents organizations from leveraging AI / ML analytics to uncover the customer, product, and operational insights (propensities) buried in the data, that then can be used to optimize the organization's business and operational use cases (see Figure 1). Then on my early morning jog, it became clear to me what organizations specifically need to do to cross the analytics chasm. And the key to successfully crossing the analytics chasm is found in my favorite and maybe my most powerful concept – the Big Data Business Model Maturity Index (see Figure 2).

Artificial intelligences (AI) are increasingly being embodied and embedded in the world to carry out tasks and support decision-making with and for people. Robots, recommender systems, voice assistants, virtual humans - do these disparate types of embodied AI have something in common? Here we show how they can manifest as "socially embodied AI." We define this as the state that embodied AI "circumstantially" take on within interactive contexts when perceived as both social and agentic by people. We offer a working ontology that describes how embodied AI can dynamically transition into socially embodied AI. We propose an ontological heuristic for describing the threshold: the Tepper line. We reinforce our theoretical work with expert insights from a card sort workshop. We end with two case studies to illustrate the dynamic and contextual nature of this heuristic.

Argo AI is expanding its presence in the Strip District. The self-driving car startup is taking more space in the Riverfront West building in the 3 Crossings development in the Strip. Argo moved its headquarters to the five-story building in 2018. It currently occupies floors four and five, as well as part of the first floor. The additional space will be on the third floor, which it will be sharing with Oxford Development Co., the developer behind 3 Crossings.