Probing for consciousness in machines

The paper proposes the design approach as a blueprint for building a sentient artificial agent capable of exhibiting humanlike attributions of consciousness. The paper also considers whether if such an artificial agent is ever built, how it will be indistinguishable from a human being? Well, it is glowingly evident that the evolution of artificial intelligence is guided by us, humans, whose own mental evolution have been shaped by the passing years in the course of the phenomenology of adaptation and survival (Darwinian). Yet, the evolution of synthetic minds powered by artificial cognition seems to be quite fast. Yes, the artificial mind in robots, if we accept the analogy ‘mind’ in its fullest sense, that day is not very far when the mental embodiment of consciousness in machines would become reality. But prior to such a feat becoming reality, rhetoric debates have been taking shape as of, how to decode and cipher consciousness in machines, a phenomenon considered as often as ‘nonentity’, then, what would be the true essence of such an artificial consciousness? This paper discusses these aspects and attempts to throw some new light on the design and developmental aspects of artificial consciousness.

The progress in the machine consciousness research field has to be assessed in terms of the features demonstrated by the new models and implementations currently being designed. In this paper, we focus on the functional aspects of consciousness and propose the application of a revision of ConsScale — a biologically inspired scale for measuring cognitive development in artificial agents — in order to assess the cognitive capabilities of machine consciousness implementations. We argue that the progress in the implementation of consciousness in artificial agents can be assessed by looking at how key cognitive abilities associated to consciousness are integrated within artificial systems. Specifically, we characterize ConsScale as a partially ordered set and propose a particular dependency hierarchy for cognitive skills. Associated to that hierarchy a graphical representation of the cognitive profile of an artificial agent is presented as a helpful analytic tool. The proposed evaluation schema is discussed and applied to a number of significant machine consciousness models and implementations. Finally, the possibility of generating qualia and phenomenological states in machines is discussed in the context of the proposed analysis.

Machine ethics has received increasing attention over the past few years because of the need to ensure safe and reliable artificial intelligence (AI). The two dominantly used theories in machine ethics are deontological and utilitarian ethics. Virtue ethics, on the other hand, has often been mentioned as an alternative ethical theory. While this interesting approach has certain advantages over popular ethical theories, little effort has been put into engineering artificial virtuous agents due to challenges in their formalization, codifiability, and the resolution of ethical dilemmas to train virtuous agents. We propose to bridge this gap by using role-playing games riddled with moral dilemmas. There are several such games in existence, such as Papers, Please and Life is Strange, where the main character encounters situations where they must choose the right course of action by giving up something else dear to them. We draw inspiration from such games to show how a systemic role-playing game can be designed to develop virtues within an artificial agent. Using modern day AI techniques, such as affinity-based reinforcement learning and explainable AI, we motivate the implementation of virtuous agents that play such role-playing games, and the examination of their decisions through a virtue ethical lens. The development of such agents and environments is a first step towards practically formalizing and demonstrating the value of virtue ethics in the development of ethical agents.

Abstract: Research links a progressive decrement in episodic memory to deficits in judgment and decision making in aging. In social decision making, decrements in episodic memory contribute to suboptimal decisions among older adults. As artificial agents – such as humanoid robots, artificial intelligence (AI)-powered deepfakes, and chatbots – become increasingly present in human society, they can both benefit (e.g., combat loneliness) and harm (e.g., defraud) older adults. Understanding how older adults make social judgments and decisions involving artificial agents is crucial for research and policy making. Nevertheless, it remains poorly understood how episodic memory deficits influence social judgments and decisions involving artificial agents, and how this relation changes with age. This review will bridge this gap by applying a developmental model to explain how memory influences social judgments and decisions involving artificial agents in late adulthood. Evidence suggests that older adults increasingly rely on gist-based processing, which may explain their greater preference for humanoid robots compared with younger adults. We will also discuss the distinct challenges deepfakes pose for older adults’ trust-related decision making. In closing, we will discuss the implications of age-related memory changes for social judgments and decisions involving disembodied artificial agents, such as chatbots.

Creativity is intrinsic to Humanities and STEM disciplines. In the activities of artists and engineers, for example, an attempt is made to bring something new into the world through counterfactual thinking. However, creativity in these disciplines is distinguished by differences in motivations and constraints. For example, engineers typically direct their creativity toward building solutions to practical problems, whereas the outcomes of artistic creativity, which are largely useless to practical purposes, aspire to enrich the world aesthetically and conceptually. In this essay, an artist (DHS) and a roboticist (GS) engage in a cross-disciplinary conceptual analysis of the creative problem of artificial consciousness in a robot, expressing the counterfactual thinking necessitated by the problem, as well as disciplinary differences in motivations, constraints, and applications. We especially deal with the question of why one would build an artificial consciousness and we consider how an illusionist theory of consciousness alters prominent ethical debates on synthetic consciousness. We discuss theories of consciousness and their applicability to synthetic consciousness. We discuss practical approaches to implementing artificial consciousness in a robot and conclude by considering the role of creativity in the project of developing an artificial consciousness.

Reinforcement Learning (RL) applications in real-world scenarios must prioritize safety and reliability, which impose strict constraints on agent behavior. Model-based RL leverages predictive world models for action planning and policy optimization, but inherent model inaccuracies can lead to catastrophic failures in safety-critical settings. We propose a novel model-based RL framework that jointly optimizes task performance and safety. To address world model errors, our method incorporates an adaptive mechanism that dynamically switches between model-based planning and direct policy execution. We resolve the objective mismatch problem of traditional model-based approaches using an implicit world model. Furthermore, our framework employs dynamic safety thresholds that adapt to the agent’s evolving capabilities, consistently selecting actions that surpass safe policy suggestions in both performance and safety. Experiments demonstrate significant improvements over non-adaptive methods, showing that our approach optimizes safety and performance simultaneously rather than merely meeting minimum safety requirements. The proposed framework achieves robust performance on diverse safety-critical continuous control tasks, outperforming existing methods.

Meta-learning, social cognition and consciousness in brains and machines

In computer science, reinforcement learning is a powerful framework with which artificial agents can learn to maximize their performance for any given Markov decision process (MDP). Advances over the last decade, in combination with deep neural networks, have enjoyed performance advantages over humans in many difficult task settings. However, such frameworks perform far less favorably when evaluated in their ability to generalize or transfer representations across different tasks. Existing algorithms that facilitate transfer typically are limited to cases in which the transition function or the optimal policy is portable to new contexts, but achieving “deep transfer” characteristic of human behavior has been elusive. Such transfer typically requires discovery of abstractions that permit analogical reuse of previously learned representations to superficially distinct tasks. Here, we demonstrate that abstractions that minimize error in predictions of reward outcomes generalize across tasks with different transition and reward functions. Such reward-predictive representations compress the state space of a task into a lower dimensional representation by combining states that are equivalent in terms of both the transition and reward functions. Because only state equivalences are considered, the resulting state representation is not tied to the transition and reward functions themselves and thus generalizes across tasks with different reward and transition functions. These results contrast with those using abstractions that myopically maximize reward in any given MDP and motivate further experiments in humans and animals to investigate if neural and cognitive systems involved in state representation perform abstractions that facilitate such equivalence relations.

Applying reinforcement learning to dialogue policy learning requires prohibitively large rounds of human-machine interactions. To improve the learning performance, the Deep Dyna-Q framework with a world model that imitates real users is widely used in recent years. Unfortunately, how to build an effective world model and how to evaluate the experiences generated by the world model efficiently have not been well studied. In order to further improve the effectiveness and efficiency of dialogue policy learning, we present a novel Gaussian Process based Deep Dyna-Q approach in this paper. The Gaussian Process model, which is analytically tractable and fits for small-sample problems, is introduced to build the world model. In addition, we design a highly efficient Kullback-Leibler divergence based discriminator to evaluate the quality of experiences generated by the world model. Extensive experiments validate the effectiveness and robustness of our proposed approach. The task-completion success rate can be improved by about 20\% with fewer human-machine interactions.

I compare a ‘realist’ with a ‘social–relational’ perspective on our judgments of the moral status of artificial agents (AAs). I develop a realist position according to which the moral status of a being—particularly in relation to moral patiency attribution—is closely bound up with that being’s ability to experience states of conscious satisfaction or suffering (CSS). For a realist, both moral status and experiential capacity are objective properties of agents. A social relationist denies the existence of any such objective properties in the case of either moral status or consciousness, suggesting that the determination of such properties rests solely upon social attribution or consensus. A wide variety of social interactions between us and various kinds of artificial agent will no doubt proliferate in future generations, and the social–relational view may well be right that the appearance of CSS features in such artificial beings will make moral role attribution socially prevalent in human–AA relations. But there is still the question of what actual CSS states a given AA is capable of undergoing, independently of the appearances. This is not just a matter of changes in the structure of social existence that seem inevitable as human–AA interaction becomes more prevalent. The social world is itself enabled and constrained by the physical world, and by the biological features of living social participants. Properties analogous to certain key features in biological CSS are what need to be present for nonbiological CSS. Working out the details of such features will be an objective scientific inquiry.

Artificial intelligence of large ensembles of intelligent agents in terms of computing power surpasses human intelligence. He is capable of artificial thinking and understanding. Giant ensembles of intellectual agents with artificial consciousness and intelligence are able, for the results set by the person necessary for him, to find solutions for their obtaining on the basis of professional competence and experience accumulation. The professional competence of artificial intelligence is the ability to use technologies, including computer vision, natural language processing, speech recognition and synthesis, intelligent decision support, as well as the use of synergistic methods, functional self-organizing methods and utility and preference criteria. For artificial intelligence, the functional organization of the system is important. The functionalism of artificial intelligence does not depend on its carrier, allows many methods of its functional implementation, as well as to form the completeness of its functions. Giant ensembles of intellectual agents with a full set of functions gradually and flexibly form events into solutions or rational behavior to obtain a given necessary result. Intelligent artificial intelligence has psychological, analytical, research, language, professional and behavioral hybrid competencies. Each competence is exercised by an intelligent agent with a competent functional professional manner. Intelligent agents form an ensemble with intelligent ethical artificial intelligence. Functional harmonious self-organization of the interaction of intellectual agents in different environments is carried out on the basis of data of a specific environment obtained by analytical competent intellectual agents. For each set of functions and smart hybrid competencies of an intelligent ensemble, there is a critical importance for the number of its intellectual agents capable of synergistic self-organization of interaction.

Inquiry into the field of artificial intelligence (machines) and its potential to develop consciousness is presented in this study. This investigation explores the complex issues surrounding machine consciousness at the nexus of AI, neuroscience, and philosophy as we delve into the fascinating world of artificial intelligence (AI) and investigate the intriguing question: are machines on the verge of becoming conscious beings? The study considers the likelihood of machines displaying self-awareness and the implications thereof through an analysis of the current state of AI and its limitations. However, with advancements in machine learning and cognitive computing, AI systems have made significant strides in emulating human-like behavior and decision-making. Furthermore, the emergence of machine consciousness raises questions about the blending of human and artificial intelligence, and ethical considerations are also considered. The study provides a glimpse into a multidisciplinary investigation that questions accepted theories of consciousness, tests the limits of what is possible with technology, and do these advancements signify a potential breakthrough in machine consciousness.

The practical use of the concepts of “reinforcement” and “punishment” have evolved in two distinct scientific domains. One domain is the psychological and biological sciences à la behaviorists of the 19th and 20th centuries. The second domain is computer science à la artificial intelligence researchers of the 20th century. Behavior scientists generally focus on understanding how biological organisms do behave as a function of their interaction with the environment. Computer scientists within artificial intelligence generally focus on how artificial agents ought to behave in specific environments to maximize cumulative reward or minimize cumulative punishment. This chapter outlines the unique scientific histories of these two areas of reinforcement learning, the commonalities and differences between these two disciplines, and how researchers might combine knowledge from each area to advance their research agenda.

The rise of Artificial Intelligence (AI) has produced prophets and prophecies announcing that the age of artificial consciousness is near. Not only does the mere idea that any machine could ever possess the full potential of human consciousness suggest that AI could replace the role of God in the future, it also puts into question the fundamental human right to freedom and dignity. This position paper takes the stand that, in the light of all we currently know about brain evolution and the never-stopping formation of adaptive neural circuitry for learning, memory, decision making and, ultimately, fully conscious reasoning and creativity in the human species, the idea of an artificial consciousness appears misconceived. The paper highlights some of the major reasons why. While awareness to external stimuli for processes such as perception, recognition, and operational problem solving is under the direct control of functionally specific brain networks associated with sensory and cognitive functions across animal species, consciousness is a unique property of the human mind. Potentiated by brain evolution, consciousness has come to be when humans became able to represent, and reflect on, the Self in relation to past, present and future, and to project these representations into possible worlds by drawing and other forms of conceptual and creative expression. Epigenetically determined, shaped by experience, capable of representing real and non-real world states, consciousness is enabled by context-dependent adaptive brain circuits that have evolved on the grounds of self-organizing functional interactions at different levels of integration in a from-local-to global functional brain design. The evolution of the latter being continuous, the limits of consciousness are unpredictable. If cracking the computational code to human consciousness were possible, the resulting algorithms would have to be able to generate temporal activity patterns simulating long-distance signal reverberation across the brain and the de-correlation of spatial signal contents from their temporal signatures. In the light of scientific evidence for complex interactions between implicit (non-conscious) and explicit (conscious) representations in learning, memory, and the construction of conscious representation, the code would have to be capable of making all implicit processing explicit. Algorithms would have to be capable of a progressive, less and less arbitrary selection of temporal activity patterns in a continuously developing neural network structure akin to that of the human brain, from synapses to higher cognitive functional integration. The code would have to possess the self-organizing capacities that generate the brain signatures of phenomenal consciousness. In the biological brain, consolidation or extinction of these temporal brain signatures is driven by external event probabilities according to the principles of Hebbian learning. Consciousness is constantly fed by such learning, capable of generating stable representations despite an incommensurable amount of variability in input data, across time and across individuals, for life-long integration of experience data. Artificial consciousness would require probabilistic adaptive computations capable of emulating all the dynamics of human learning and memory that enable human intelligence and creativity. No AI is likely to ever have such potential.

The Best Laid Plans: Computational Principles of Anterior Cingulate Cortex.