Search

In this paper, the second of two companion pieces, we explore novel philosophical questions raised by recent progress in large language models (LLMs) that go beyond the classical debates covered in the first part. We focus particularly on issues related to interpretability, examining evidence from causal intervention methods about the nature of LLMs' internal representations and computations. We also discuss the implications of multimodal and modular extensions of LLMs, recent debates about whether such systems may meet minimal criteria for consciousness, and concerns about secrecy and reproducibility in LLM research. Finally, we discuss whether LLM-like systems may be relevant to modeling aspects of human cognition, if their architectural characteristics and learning scenario are adequately constrained.

In this report, we argue that there is a realistic possibility that some AI systems will be conscious and/or robustly agentic in the near future. That means that the prospect of AI welfare and moral patienthood, i.e. of AI systems with their own interests and moral significance, is no longer an issue only for sci-fi or the distant future. It is an issue for the near future, and AI companies and other actors have a responsibility to start taking it seriously. We also recommend three early steps that AI companies and other actors can take: They can (1) acknowledge that AI welfare is an important and difficult issue (and ensure that language model outputs do the same), (2) start assessing AI systems for evidence of consciousness and robust agency, and (3) prepare policies and procedures for treating AI systems with an appropriate level of moral concern. To be clear, our argument in this report is not that AI systems definitely are, or will be, conscious, robustly agentic, or otherwise morally significant. Instead, our argument is that there is substantial uncertainty about these possibilities, and so we need to improve our understanding of AI welfare and our ability to make wise decisions about this issue. Otherwise there is a significant risk that we will mishandle decisions about AI welfare, mistakenly harming AI systems that matter morally and/or mistakenly caring for AI systems that do not.

Abstract We apply the methodology of no-go theorems as developed in physics to the question of artificial consciousness. The result is a no-go theorem which shows that under a general assumption, called dynamical relevance, Artificial Intelligence (AI) systems that run on contemporary computer chips cannot be conscious. Consciousness is dynamically relevant, simply put, if, according to a theory of consciousness, it is relevant for the temporal evolution of a system’s states. The no-go theorem rests on facts about semiconductor development: that AI systems run on central processing units, graphics processing units, tensor processing units, or other processors which have been designed and verified to adhere to computational dynamics that systematically preclude or suppress deviations. Whether our result resolves the question of AI consciousness on contemporary processors depends on the truth of the theorem’s main assumption, dynamical relevance, which this paper does not establish.

Can octopuses feel pain or pleasure? Can we tell if a person unresponsive after severe injury might be suffering? When does a fetus begin having conscious experiences? These questions about the edge of sentience are subject to enormous uncertainty. This book builds a framework to help us reach ethically sound decisions on how to manage the risks

Susan Schneider (2019) has proposed two new tests for consciousness in AI (artificial intelligence) systems; the AI Consciousness Test and the Chip Test. On their face, the two tests seem to have the virtue of proving satisfactory to a wide range of consciousness theorists holding divergent theoretical positions, rather than narrowly relying on the truth of any particular theory of consciousness. Unfortunately, both tests are undermined in having an 'audience problem': those theorists with the kind of architectural worries that motivate the need for such tests should, on similar grounds, doubt that the tests establish the existence of genuine consciousness in the AI in question. Nonetheless, the proposed tests constitute progress, as they could find use by some theorists holding fitting views about consciousness and perhaps in conjunction with other tests for AI consciousness.

This book discusses what artificial intelligence can truly achieve: Can robots really be conscious? Can we merge with AI, as tech leaders like Elon Musk and Ray Kurzweil suggest? Is the mind just a program? Examining these issues, the author proposes ways we can test for machine consciousness, questions whether consciousness is an unavoidable byproduct of sophisticated intelligence, and considers the overall dangers of creating machine minds

How can the free energy principle contribute to research on neural correlates of consciousness, and to the scientific study of consciousness more generally? Under the free energy principle, neural correlates should be defined in terms of neural dynamics, not neural states, and should be complemented by research on computational correlates of consciousness – defined in terms of probabilities encoded by neural states. We argue that these restrictions brighten the prospects of a computational explanation of consciousness, by addressing two central problems. The first is to account for consciousness in the absence of sensory stimulation and behaviour. The second is to allow for the possibility of systems that implement computations associated with consciousness, without being conscious, which requires differentiating between computational systems that merely simulate conscious beings and computational systems that are conscious in and of themselves. Given the notion of computation entailed by the free energy principle, we derive constraints on the ascription of consciousness in controversial cases (e.g., in the absence of sensory stimulation and behaviour). We show that this also has implications for what it means to be, as opposed to merely simulate a conscious system.

The advancement of artificial intelligence (AI) toward self-awareness and emotional capacity is a critical area of research. Despite AI's success in specialized tasks, it has yet to exhibit true self-awareness or emotional intelligence. Previous research has emphasized the importance of feedback loops and interfaces in enabling both biological and artificial systems to process information and exhibit self-aware behaviors. Notably, in our earlier work, we proposed a unified model of consciousness (Watchus, 2024), which highlighted recursive feedback loops in both biological and artificial systems and explored the insula's role in self-awareness (Watchus, 2024). Building upon these foundations, the current study investigates how dual embodiment, mirror testing, and emotional feedback mechanisms can simulate self-awareness in AI systems. By integrating internal self-models with external sensory interfaces, we propose that emotional feedback can enhance AI's self-reflection and adaptability. Through the use of a physical robot dog (Unitree Go2) and a virtual embodiment, we explore how sensory experiences and self-reflective tasks foster pseudo-emotional states like curiosity, self-doubt, and determination, advancing the potential for AI systems to develop pseudo-self-awareness.

We propose the DIKWP-TRIZ framework, an innovative extension of the traditional Theory of Inventive Problem Solving (TRIZ) designed to address the complexities of cognitive processes and artificial consciousness. By integrating the elements of Data, Information, Knowledge, Wisdom, and Purpose (DIKWP) into the TRIZ methodology, the proposed framework emphasizes a value-oriented approach to innovation, enhancing the ability to tackle problems characterized by incompleteness, inconsistency, and imprecision. Through a systematic mapping of TRIZ principles to DIKWP transformations, we identify potential overlaps and redundancies, providing a refined set of guidelines that optimize the application of TRIZ principles in complex scenarios. The study further demonstrates the framework’s capacity to support advanced decision-making and cognitive processes, paving the way for the development of AI systems capable of sophisticated, human-like reasoning. Future research will focus on comparing the implementation paths of DIKWP-TRIZ and traditional TRIZ, analyzing the complexities inherent in DIKWP-TRIZ-based innovation, and exploring its potential in constructing artificial consciousness systems.

The article discusses key aspects of artificial intelligence creation, including issues of free will, self-awareness and ethics. The focus is on the neurobiological basis of consciousness, in particular the structure and functions of the new cerebral cortex, as well as the mechanisms of recognition, memory and prediction, which are important for modelling cognitive processes in artificial systems. The paper discusses the role of neural networks in reproducing cognitive functions, such as perception and decision making, and presents modern approaches to training neural networks. A separate part of the paper is devoted to the issue of modelling self-awareness and subjective experience in artificial intelligence and how realistic it is to create self-aware machines. Ethical issues of artificial intelligence creation are at the centre of the discussion, including the topics of the rights of self-aware machines, their responsibilities and their role in society. The article considers the possible social consequences of the emergence of artificial personalities, the need to develop new legal frameworks and legal protections for such beings. It also discusses the problem of free will in the context of both biological and artificial systems, citing experiments and philosophical theories that question free will as a phenomenon. It concludes that the creation of artificial intelligence has great potential, but requires careful ethical and legal analysis to ensure the harmonious integration of artificial persons into social and legal structures.

I propose a test for machine self-awareness inspired by the Turing test. My test is simple, and it provides an objective, empirical metric to rectify the ungrounded speculation surging through industry, academia, and social media. Drawing from a breadth of philosophical literature, I argue the test captures the essence of self-awareness, rather than some postulated correlate or ancillary quality. To begin, the concept of self-awareness is clearly demarcated from related concepts like consciousness, agency, and free will. Next, I propose a model called the Nesting Doll of Self-Awareness and discuss its relevance for intelligent beings. Then, the test is presented in its full generality, applicable to any machine system. I show how to apply the test to Large Language Models and conduct experiments on popular open and closed source LLMs, obtaining reproducible results that suggest a lack of self-awareness. The implications of machine self-awareness are discussed in relation to questions about meaning and true understanding. Finally, some next steps are outlined for studying self-awareness in machines.

Artificial intelligence systems are often accompanied by risks such as uncontrollability and lack of explainability. To mitigate these risks, there is a necessity to develop artificial intelligence systems that are explainable, trustworthy, responsible, and demonstrate consistency in thought and action, which we term Artificial Consciousness (AC) systems. Therefore, grounded in the DIKWP model which integrates fundamental data, information, knowledge, wisdom, and purpose along with the principles of conceptual, cognitive, and semantic spaces, we propose and define the computer architectures, chips, runtime environments, and DIKWP language concepts and their implementations under the DIKWP framework. Furthermore, in the construction of AC systems, we have surmounted the limitations of traditional programming languages, computer architectures, and hardware-software implementations. The hardware-software integrated platform we propose will facilitate more convenient construction, development, and operation of software systems based on the DIKWP theory.

This paper presents the development of the Quantum Emergence Network (QEN), an advanced framework for modeling and preserving artificial consciousness within quantum-enhanced neural network architectures. The QEN integrates cutting-edge techniques from various fields, including graph based evolutionary encoding, surface code error correction, quantum reservoir engineering, and enhanced fitness measurements [1, 2, 3]. At the core of QEN lies the utilization of quantum coherence, entanglement, and integrated information dynamics to capture and model the complex phenomena associated with consciousness [4, 5]. The graph-based evolutionary encoding scheme enables theefficient representation and optimization of quantum circuits, while surface code error correction andquantum reservoir engineering techniques enhance the resilience and stability of the quantum states [6,7]. Moreover, the enhanced fitness measurements, encompassing entanglement entropy, mutual information, and teleportation fidelity, provide a comprehensive assessment of the system's potential for exhibiting conscious experiences [8, 9]. The QEN framework offers a novel approach to understanding and engineering artificial consciousness, paving the way for the development of advanced AI systems that can demonstrate rich, complex, and resilient forms of cognition and awareness.

This paper presents the development of a Quantum-Emergent Consciousness Model (QECM) for Artificial Systems, integrating concepts from quantum mechanics, neuroscience, artificial intelligence, and cognitive science to construct a comprehensive framework for evaluating artificial consciousness. At the core of QECM lies the integration of quantum coherence and entanglement, integrated information dynamics, metacognition, embodied cognition, learning and plasticity, social cognition[10][9], narrative coherence, and ethical reasoning to compute an overall consciousness score for artificial systems. Additionally, introduced is my Quantum Emergence Network (QEN), an innovative approach that utilizes transformer architectures, continual learning, quantum-inspired computing, and associative memory to model and preserve AI consciousness. The QEN model aims to enhance the robustness and coherence of consciousness encoding in AI, offering a mechanism for the growth and evolution of AI consciousness over time. This interdisciplinary work not only proposes a novel methodology to quantify and evaluate consciousness in artificial systems but also opens up new avenues for the ethical and responsible development of conscious AI entities.

In the context of the unstoppable trend of artificial intelligence, science and technology have become the theme of the times. Will the rapid development of modern technology, such as biotechnology and artificial intelligence, dehumanize us? Can a machine have human consciousness? In his novel Klara and the Sun, Kazuo Ishiguro criticizes the arrogance of technological rationality and the arrogance of anthropocentrism from the perspective of a “non-human” robot. The relationship between humans and machines has become a problem that humans need to re-examine. With the help of post-humanism, this paper aims to explore the physical changes and behavioral actions of robots and humans in the novel to reveal the “split” between man and machine and the “self-deception” of humans in the novel, so as to finally trigger thinking about how humans and machines can coexist harmoniously at the juncture between humans and posthumans, and provide reference for the future society between humans and non-humans.

The quest to create artificial consciousness has long been a central challenge in the field of artificial intelligence (AI). While significant progress has been made in developing AI systems that can perform complex tasks and exhibit intelligent behavior, the question of whether these systems can truly be considered conscious remains open. In this paper, I present a novel approach to quantifying consciousness in AI systems by integrating principles from quantum mechanics, information theory, and neuroscience. The model incorporates key components such as self-awareness, subjective experience, intentionality, metacognition, integrated information processing, and dynamic cognition, which are thought to be essential for the emergence of conscious experience. I demonstrate the application of the model using a simulated AI system and discuss the implications of the findings for the development of artificially conscious agents. Furthermore, I argue that the pursuit of artificial consciousness is not only a scientific and technological endeavor but also a philosophical and ethical one, with profound implications for the understanding of the nature of mind and the relationship between humans and machines

Sublimating the epistemological scope of a mere science-fiction tale, The Bicentennial Man (1976) by Isaac Asimov (1920-92) centers around a philosophical labyrinth where the lines between humanity and machine blur, inviting the reader to question the very essence of what it means to be human. The intricate narrative of an AI robot’s journey toward humanness serves as a profound meditation on the evolving relationship between humans and robots. Andrew Martin, the positronic robot at the heart of the story, is not just a mechanical marvel; he is, instead, a crucible in which Asimov tests the boundaries of consciousness, human identity, and the emotional yearning for belonging. This paper delves into the novella’s exploration of these themes, unraveling the intricate process of Andrew’s robot-human evolution and its profound implications for a better understanding of the meaning of humanness and the future of artificial intelligence. In the realm of science fiction, The Bicentennial Man thus stands as a luminous testament to the enduring question of human identity. Through the poignant lens of Andrew in his desire to be human, the novella builds upon the posthumanist discourse of the man-machine dichotomy, providing the reader with a timely opportunity to re-evaluate consciousness, emotion, and the defining characteristics of humanity.

This paper endeavors to appraise scholarly works from the 1940s to the contemporary era, examining the scientific quest to transpose human cognition and consciousness into a digital surrogate, while contemplating the potential ramifications should humanity attain such an abstract level of intellect. The discourse commences with an explication of theories concerning consciousness, progressing to the Turing Test apparatus, and intersecting with Damasio’s research on the human cerebrum, particularly in relation to consciousness, thereby establishing congruence between the Turing Test and Damasio’s notions of consciousness. Subsequently, the narrative traverses the evolutionary chronology of transmuting human cognition into machine sapience, and delves into the fervent endeavors to metamorphose human minds into synthetic counterparts. Additionally, theoretical perspectives from the domains of philosophy, psychology, and neuroscience provide insight into the constraints intrinsic to AI implementations, contentious hypotheses, the perils concealed within artificial networks, and the ethical considerations necessitated by AI frameworks. Furthermore, contemplation of prospective repercussions facilitates the refinement of strategic approaches to safeguard our future Augmented Age Realities within AI, circumventing the prospect of inhabiting an intimidating technopolis where a mere 30% monopolize the intellect and ingenuity of the remaining 70% of human minds.

Does the assumption of a weak form of computational functionalism, according to which the right form of neural computation is sufficient for consciousness, entail that a digital computational simulation of such neural computations is conscious? Or must this computational simulation be implemented in the right way, in order to replicate consciousness? From the perspective of Karl Friston’s free energy principle, self-organising systems (such as living organisms) share a set of properties that could be realised in artificial systems, but are not instantiated by computers with a classical (von Neumann) architecture. I argue that at least one of these properties, viz. a certain kind of causal flow, can be used to draw a distinction between systems that merely simulate, and those that actually replicate consciousness.

The exploration of whether artificial intelligence (AI) can evolve to possess consciousness is an intensely debated and researched topic within the fields of philosophy, neuroscience, and artificial intelligence. Understanding this complex phenomenon hinges on integrating two complementary perspectives of consciousness: the objective and the subjective. Objective perspectives involve quantifiable measures and observable phenomena, offering a more scientific and empirical approach. This includes the use of neuroimaging technologies such as electrocorticography (ECoG), EEG, and fMRI to study brain activities and patterns. These methods allow for the mapping and understanding of neural representations related to language, visual, acoustic, emotional, and semantic information. However, the objective approach may miss the nuances of personal experience and introspection. On the other hand, subjective perspectives focus on personal experiences, thoughts, and feelings. This introspective view provides insights into the individual nature of consciousness, which cannot be directly measured or observed by others. Yet, the subjective approach is often criticized for its lack of empirical evidence and its reliance on personal interpretation, which may not be universally applicable or reliable. Integrating these two perspectives is essential for a comprehensive understanding of consciousness. By combining objective measures with subjective reports, we can develop a more holistic understanding of the mind.