GPU Domain Specialization via Composable On-Package Architecture

As GPUs scale their low-precision matrix math throughput to boost deep learning (DL) performance, they upset the balance between math throughput and memory system capabilities. We demonstrate that a converged GPU design trying to address diverging architectural requirements between FP32 (or larger)-based HPC and FP16 (or smaller)-based DL workloads results in sub-optimal configurations for either of the application domains. We argue that a C omposable O n- PA ckage GPU (COPA-GPU) architecture to provide domain-specialized GPU products is the most practical solution to these diverging requirements. A COPA-GPU leverages multi-chip-module disaggregation to support maximal design reuse, along with memory system specialization per application domain. We show how a COPA-GPU enables DL-specialized products by modular augmentation of the baseline GPU architecture with up to 4× higher off-die bandwidth, 32× larger on-package cache, and 2.3× higher DRAM bandwidth and capacity, while conveniently supporting scaled-down HPC-oriented designs. This work explores the microarchitectural design necessary to enable composable GPUs and evaluates the benefits composability can provide to HPC, DL training, and DL inference. We show that when compared to a converged GPU design, a DL-optimized COPA-GPU featuring a combination of 16× larger cache capacity and 1.6× higher DRAM bandwidth scales per-GPU training and inference performance by 31% and 35%, respectively, and reduces the number of GPU instances by 50% in scale-out training scenarios.

"I Felt a Little Crazy Following a 'Doll'"

Virtual Influencers (VIs) are computer-generated characters, many of which are often visually indistinguishable from humans and interact with the world in the first-person perspective as social media influencers. They are gaining popularity by creating content in various areas, including fashion, music, art, sports, games, environmental sustainability, and mental health. Marketing firms and brands increasingly use them to capitalise on their millions of followers. Yet, little is known about what prompts people to engage with these digital beings. In this paper, we present our interview study with online users who followed different VIs on Instagram beyond the fashion application domain. Our findings show that the followers are attracted to VIs due to a unique mixture of visual appeal, sense of mystery, and creative storytelling that sets VI content apart from that of real human influencers. Specifically, VI content enables digital artists and content creators by removing the constraints of bodies and physical features. The followers not only perceived VIs' rising popularity in commercial industries, but also are supportive of VI involvement in non-commercial causes and campaigns. However, followers are reluctant to attribute trustworthiness to VIs in general though they display trust in limited domains, e.g., technology, music, games, and art. This research highlights VI's potential as innovative digital content, carrying influence and employing more varied creators, an appeal that could be harnessed by diverse industries and also by public interest organisations.

What Are the Boundaries to the Expansion of Digital Labour Platforms? Understanding Uberization through a Cognitive Sustainability Lens

While digital labour platforms are booming, their ability to constitute a sustainable alternative to the managerial firm and to salaried work is questionable. To date, this debate has been approached mainly from legal or political angles, and the organizational sustainability of such platforms remains underexplored. We respond to calls to study more specifically the cognitive capabilities of platforms by mobilizing knowledge-based theories of the firm. We contribute to the literature in three ways: (1) we introduce the concept of ‘cognitive sustainability’, which we define as the capacity to ensure the integration, conservation and creation of knowledge; (2) we develop a set of propositions aimed at identifying the activities that platforms are most likely to carry out in a cognitively sustainable way; (3) we argue for the possibility of an increased hybridization of digital labour platforms to perform complex activities. Mobilizing knowledge-based theories of the firm to explore new objects such as platforms and taking such hybridization processes into account adds to this body of literature by extending its application domain and taking a more dynamic perspective.

Agent-Based Modelling and Simulation with Domain-Specific Languages

Conducting simulation studies within a model-based framework is a complex process, in which many different concerns must be considered. Central tasks include the specification of the simulation model, the execution of simulation runs, the conduction of systematic simulation experiments, and the management and documentation of the model’s context. In this chapter, we look into how these concerns can be separated and handled by applying domain-specific languages (DSLs), that is, languages that are tailored to specific tasks in a specific application domain. We demonstrate and discuss the features of the approach by using the modelling language ML3, the experiment specification language SESSL, and PROV, a graph-based standard to describe the provenance information underlying the multi-stage process of model development.

An approach for Industrie 4.0-compliant and data-sovereign Digital Twins

Data sharing between enterprises requires both interoperability and data sovereignty. In the application domain of industrial production an integrated approach is required that encompasses standards and technologies of both Industrie 4.0 and the International Data Spaces (IDS). This paper describes how to combine them for the concept of Digital Twins following the architectural framework given in ISO DIS 23247. Furthermore, an implementation approach is described relying upon the Fraunhofer Advanced AAS Tools for Digital Twins (FA³ST). The resulting architectural approach may be combined with further open manufacturing standards, and may be applied for data analytics and the engineering of AI-based systems.

Proposal of Algorithm for Automated Chromosomal Abnormality Detection

The article considers the problem of automatic chromosome abnormalities recognition, using images of chromosomes as an input. This paper’s scope includes overview of application domain and analysis of existing solutions. A high-level algorithm for chromosome abnormalities recognition automation is proposed, and a proof-of-concept application is built on top of the algorithm.

What Metadata? Defining Different Types of Digital Assets as Application Targets of Metadata in Clinical Research Informatics

The term ‘metadata’ is mentioned in every one of the FAIR principles. Metadata is without question important for findability, accessibility, and reusability, but essential for interoperability. Standardized schemas have been developed by various stakeholders for decades, but too rarely come to practical use. The reason for this is that the application domain is not clearly understood. In many bio-medical research projects, the need for metadata is recognized at some point, but there is not only a lack of overview of existing standards, but also a lack of correct assessment of what individual metadata schemas were actually made for. This paper differentiates different application scenarios for metadata in clinical research.

RADAR: Resilient Application for Dependable Aided Reporting

Many organizations must produce many reports for various reasons. Although this activity could appear simple to carry out, this fact is not at all true: indeed, generating reports requires the collection of possibly large and heterogeneous data sets. Furthermore, different professional figures are involved in the process, possibly with different skills (database technicians, domain experts, employees): the lack of common knowledge and of a unifying framework significantly obstructs the effective and efficient definition and continuous generation of reports. This paper presents a novel framework named RADAR, which is the acronym for “Resilient Application for Dependable Aided Reporting”: the framework has been devised to be a ”bridge” between data and employees in charge of generating reports. Specifically, it builds a common knowledge base in which database administrators and domain experts describe their knowledge about the application domain and the gathered data; this knowledge can be browsed by employees to find out the relevant data to aggregate and insert into reports, while designing report layouts; the framework assists the overall process from data definition to report generation. The paper presents the application scenario and the vision by means of a running example, defines the data model and presents the architecture of the framework.

A computational model of argumentation schemes for multi-agent systems

There are many benefits of using argumentation-based techniques in multi-agent systems, as clearly shown in the literature. Such benefits come not only from the expressiveness that argumentation-based techniques bring to agent communication but also from the reasoning and decision-making capabilities under conditions of conflicting and uncertain information that argumentation enables for autonomous agents. When developing multi-agent applications in which argumentation will be used to improve agent communication and reasoning, argumentation schemes (reasoning patterns for argumentation) are useful in addressing the requirements of the application domain in regards to argumentation (e.g., defining the scope in which argumentation will be used by agents in that particular application). In this work, we propose an argumentation framework that takes into account the particular structure of argumentation schemes at its core. This paper formally defines such a framework and experimentally evaluates its implementation for both argumentation-based reasoning and dialogues.