by Theodore Patkos (ICS-FORTH) and Zsolt Viharos (SZTAKI)
The wave of popularity of modern Artificial Intelligence (AI) systems is creating well-justified expectations that its application to diverse domains will lead to even bigger advancements in the future. At the same time, there is an open debate in the research community regarding the limitations of existing methods, and whether the current success can scale up to a broader spectrum of problems than those current AI is focusing on. Especially considering that modern intelligent systems are affecting our everyday lives at an increasing pace, the request for future intelligent machines is to exhibit capabilities that are not only effective, but also closer to human intuition and intellect.
In order achieve a reliable human-machine symbiotic collaboration, AI needs to make progress on skills that humans excel in. These skills include: a general understanding of how the world works; the exploitation of common sense knowledge that is hidden, yet pervasive, in the majority of human-to-human interactions; the ability to understand other people’s intentions and to comply with social norms and values, or the ability to explain with grounded justification their decisions. The explainability of AI models is one of the crucial challenges in today’s machine learning solutions, in order to help designers and users understand those more-often-experienced cases, when AI systems significantly outperform human capabilities.
Cognitive AI intends to augment intelligent machines with human-like cognition. It builds on the advancements of modern, data-driven AI technologies, but also calls for progress in symbolic, knowledge-based methods, in order to enable machines to learn from how humans create rich cognitive models about the world they live in, and how they ascribe mental states to themselves and others, such as beliefs, intentions, emotions, perceptions, even thoughts. Cognitive AI will also help intelligent systems engage more smoothly in social interactions, accomplish collaborative tasks, and in general broaden their intelligence and the spectrum of problems they can tackle.
The field of collaborative robots (cobots) is a characteristic case, where Cognitive AI is expected to achieve broad impact. Cobots are receiving continuously increasing roles and benefits in social life and in the industrial sectors as well. This technical equipment shares the field of activities with human beings in space and/or time and, building on tight human-robot collaboration, it results in additional value-added processes beyond the summarised outputs of these two separate contributors. Coexistence, cooperation, sequential and responsive collaboration form the different levels of the common activities, in each of which their cross-understanding and harmonisation is an essential feature to be realised and exploited, giving rise to various novel challenges for science.
As such, Cognitive AI lies at the intersection of both Computer and Cognitive Sciences, crossing through a broad range of research fields, ranging from Robotics, and Human-Robot interaction and collaboration, to aspects of Computer Vision, such as scene understanding, or to the modelling and automation of Human-Machine social encounters and argumentative dialogues. The current ERCIM News special theme presents progress on all those matters, showcasing remarkable achievements in the industry, but also discussing highly interesting applications in everyday human life and in art.
Next, we provide a short overview on the articles included under the topic of Cognitive AI and Cobots, grouping them according to their thematic similarity.
AI in Manufacturing
Papoutsakis and Pateraki describe a vision-based framework for real-time monitoring of 3D human motion and classification of ergonomic work postures and risk assessment during assembly tasks under real conditions, in the car manufacturing industry. Similarly, Anagnostopoulos et al. present an AI-augmented multi-stereo camera system that collects ergonomic information of human operators in their working space, monitoring diverse anthropometric characteristics, in order to adjust the behaviour of robots working in the same space, accordingly. Vento et al., in addition to visual information, also try to recognise speech commands, in order to achieve a more natural human-machine interaction in a cooperative, assembly line, industrial environment.
Models for Improving Human-Machine Collaboration in Industry
In order to achieve reliable human-machine symbiotic collaboration, Calabrò and Marchetti discuss a desktop application for managing domain-specific protocols that can verify and validate a cobot's quality properties, such as safety and security. Behery et al. propose a generalisation of “Behaviour Trees” as a modelling framework for industrial human-robot collaboration. This representation enhances safety, explainability, and verification, but also modification and exchange of production processes. Karvonen and Saariluoma propose human digital twins as a conceptual design tool for a more human-centric approach to future industrial design. It is based on cognitive mimetics, a methodology for constructing technologies by mimicking human intelligent information processing. Dagioglou and Karkaletsis investigate how humans perceive and experience the collaboration with AI-embodied agents, by means of a collaborative learning testbed that uses objective and subjective measures to rate the human-machine interaction.
Argument Exchange in Human-Machine Interaction
Michael proposes a structured, dialectical model for teaching an AI entity, through the exchange of arguments and counterarguments. Kakas goes one step further and presents a system that implements Cognitive Machine Argumentation to investigate the effect of machine explanations on human reasoning.
AI in Arts
Thomay et al. apply Cognitive AI in multimodal arts installations to create interactive systems that learn from their visitors, improving not only gesture recognition, but also usability and user experience. Cserteg et al. combine Cognitive AI with Robotics to develop a portrait-drawing robot, which takes the picture of a person and, after a series of processing and execution steps, delivers a drawing with the person’s facial features as a set of lines drawn by an average industrial robot.
Cognitive AI in Everyday Life
Hung stresses the value of social encounters for humans and explores ways of measuring the quality of experience in social encounters through the interpretation of body language. Catricalà et al. (page 28) focus on the benefits that human-robot interaction can have on older adults, and discuss how serious games, based on and exploiting users’ memories, can help them maintain their cognitive functional level. Hoang et al. (page 30) provide insights into a novel method for recognising various types of activities of humans performing everyday or, in some cases, professional tasks.
Theodore Patkos, ICS-FORTH, Greece
Zsolt Viharos, SZTAKI, Hungary