Experimental Detection of Universal Critical Signatures Associated with Conscious States Across Biological and Artificial Systems

The challenges of lifelong learning in biological and artificial systems.

The computational significance of consciousness is an important and potentially more tractable research theme than the hard problem of consciousness, as one could look at the correlation of consciousness and computational capacities through, e.g., algorithmic or complexity analyses. In the literature, consciousness is defined as what it is like to be an agent (i.e., a human or a bat), with phenomenal properties, such as qualia, intentionality, and self-awareness. The absence of these properties would be termed "unconscious." The recent success of large language models (LLMs), such as ChatGPT, has raised new questions about the computational significance of human conscious processing. Although instances from biological systems would typically suggest a robust correlation between intelligence and consciousness, certain states of consciousness seem to exist without manifest existence of intelligence. On the other hand, AI systems seem to exhibit intelligence without consciousness. These instances seem to suggest possible dissociations between consciousness and intelligence in natural and artificial systems. Here, I review some salient ideas about the computational significance of human conscious processes and identify several cognitive domains potentially unique to consciousness, such as flexible attention modulation, robust handling of new contexts, choice and decision making, cognition reflecting a wide spectrum of sensory information in an integrated manner, and finally embodied cognition, which might involve unconscious processes as well. Compared to such cognitive tasks, characterized by flexible and ad hoc judgments and choices, adequately acquired knowledge and skills are typically processed unconsciously in humans, consistent with the view that computation exhibited by LLMs, which are pretrained on a large dataset, could in principle be processed without consciousness, although conversations in humans are typically done consciously, with awareness of auditory qualia as well as the semantics of what are being said. I discuss the theoretically and practically important issue of separating computations, which need to be conducted consciously from those which could be done unconsciously, in areas, such as perception, language, and driving. I propose conscious supremacy as a concept analogous to quantum supremacy, which would help identify computations possibly unique to consciousness in biologically practical time and resource limits. I explore possible mechanisms supporting the hypothetical conscious supremacy. Finally, I discuss the relevance of issues covered here for AI alignment, where computations of AI and humans need to be aligned.

Learning from sensory patterns associated with different kinds of sensors is paramount for biological systems, as it permits them to cope with complex environments where events rarely appear twice in the same way. In this paper we want to investigate how perceptual categories formed in one modality can be transferred to another modality in biological and artificial systems. We first present a study on Mongolian gerbils that show clear evidence of transfer of knowledge for a perceptual category from the auditory modality to the visual modality. We then introduce an algorithm that mimics the behavior of the rodents within the online learning framework. Experiments on simulated data produced promising results, showing the pertinence of our approach.

This paper explores some of the potential connections between natural and artificial intelligence and natural and artificial consciousness. In humans we use batteries of tests to indirectly measure intelligence. This approach breaks down when we try to apply it to radically different animals and to the many varieties of artificial intelligence. To address this issue people are starting to develop algorithms that can measure intelligence in any type of system. Progress is also being made in the scientific study of consciousness: we can neutralize the philosophical problems, we have data about the neural correlates and we have some idea about how we can develop mathematical theories that can map between physical and conscious states. While intelligence is a purely functional property of a system, there are good reasons for thinking that consciousness is linked to particular spatiotemporal patterns in specific physical materials. This paper outlines some of the weak inferences that can be made about the relationships between intelligence and consciousness in natural and artificial systems. To make real scientific progress we need to develop practical universal measures of intelligence and mathematical theories of consciousness that can reliably map between physical and conscious states.

EDITORIAL article Front. Neurorobot., 13 September 2013 https://doi.org/10.3389/fnbot.2013.00013

Twenty-first century is predicted to be an age of humanoids that would be difficult to discern from human beings. Replicating human structure, systems and functions are challenging tasks. Some progress is made in developing artificial intelligence systems, by mimicking the biological systems in humans. However, a holistic approach is needed towards developing all the human-systems, in tandem. In this study, the biological and artificial systems are compared. The study focuses upon the need to develop more biomedically integrated components, for better replication of the human systems. A prioritization of the systems in humanoids, as compared to humans, is also proposed.

In this project we studied both natural photosynthetic antenna complexes and various artificial systems (e.g. chlorophyll (Chl) trefoils) using high resolution hole-burning (HB) spectroscopy and excitonic calculations. Results obtained provided more insight into the electronic (excitonic) structure, inhomogeneity, electron-phonon coupling strength, vibrational frequencies, and excitation energy (or electron) transfer (EET) processes in several antennas and reaction centers. For example, our recent work provided important constraints and parameters for more advanced excitonic calculations of CP43, CP47, and PSII core complexes. Improved theoretical description of HB spectra for various model systems offers new insight into the excitonic structure and composition of low-energy absorption traps in very several antenna protein complexes and reaction centers. We anticipate that better understanding of HB spectra obtained for various photosynthetic complexes and their simultaneous fits with other optical spectra (i.e. absorption, emission, and circular dichroism spectra) provides more insight into the underlying electronic structures of these important biological systems. Our recent progress provides a necessary framework for probing the electronic structure of these systems via Hole Burning Spectroscopy. For example, we have shown that the theoretical description of non-resonant holes is more restrictive (in terms of possible site energies) than those of absorption and emission spectra. We have demonstrated that simultaneous description of linear optical spectra along with HB spectra provides more realistic site energies. We have also developed new algorithms to describe both nonresonant and resonant hole-burn spectra using more advanced Redfield theory. Simultaneous description of various optical spectra for complex biological system, e.g. artificial antenna systems, FMO protein complexes, water soluble protein complexes, and various mutants of reaction centers continues; this work is supported by the new DOE BES grant.

The rapid advances in the capabilities of Large Language Models (LLMs) have galvanised public and scientific debates over whether artificial systems might one day be conscious. Prevailing optimism is often grounded in computational functionalism: the assumption that consciousness is determined solely by the right pattern of information processing, independent of the physical substrate. Opposing this, biological naturalism insists that conscious experience is fundamentally dependent on the concrete physical processes of living systems. Despite the centrality of these positions to the artificial consciousness debate, there is currently no coherent framework that explains how biological computation differs from digital computation, and why this difference might matter for consciousness. Here, we argue that the absence of consciousness in artificial systems is not merely due to missing functional organisation but reflects a deeper divide between digital and biological modes of computation and the dynamico-structural dependencies of living organisms. Specifically, we propose that biological systems support conscious processing because they (i) instantiate scale-inseparable, substrate-dependent multiscale processing as a metabolic optimisation strategy, and (ii) alongside discrete computations, they perform continuous-valued computations due to the very nature of the fluidic substrate from which they are composed. These features - scale inseparability and hybrid computations - are not peripheral, but essential to the brain's mode of computation. In light of these differences, we outline the foundational principles of a biological theory of computation and explain why current artificial intelligence systems are unlikely to replicate conscious processing as it arises in biology.

Bioelectronic medicine is driving the need to design low-power circuits for interfacing biological neurons to electronic neural processing systems, and for implementing real-time close-loop interactions with the biological tissue. This interaction would benefit from congruent features between the biological and artificial systems, such as their working frequency and temporal dynamics. Neuromorphic engineering provides design solutions for building circuits capable of emulating biological neural processing systems faithfully. However, very few, albeit notable, attempts have been made so far to provide accurate models of action potential generation mechanisms with time-constants and dynamics that resemble those of real neurons. This paper presents a design of a silicon neuron, based on a generalized Hodgkin-Huxley model with programmable slopes for each ion channel model, that provide a robust method for matching accurately the silicon neural dynamics to those of target neuron types in biological systems. The parameters of the ion channel dynamics are controlled by a circuit comprising multiple Differential Pairs. This can be used to shape the membrane voltage profile of the silicon neuron. We use this feature to emulate biological neurons involved in respiratory Central Pattern Generator responsible for the stimulation of the vagus nerve for the activation of the heart chamber pacing. The novelty introduced in our approach is to provide a step further toward the development of a silicon neuron able to reproduce the response of biological cells and to interact with them in real-time, with the aim to design low power Brain-Machine-Interface.

Biological and artificial evolutionary systems exhibit varying degrees of evolvability and different rates of evolution. Such quantities can be affected by various factors. Here, we review some evolutionary mechanisms and discuss new developments in biology that can potentially improve evolvability or accelerate evolution in artificial systems. Biological notions are discussed to the degree they correspond to notions in Evolutionary Computation. We hope that the findings put forward here can be used to design computational models of evolution that produce significant gains in evolvability and evolutionary speed.

In the continuing quest for operational definitions of information, this paper analyzes samples of data and information measurements used in biological systems at the genetic, cellular, and organ levels. The transduction of energy into information in biological systems yields information in the form of genetic code words (condons), bit sequences, analog codes, and paired chemical markers. Pragmatic information measures applicable to many artificial systems (robots, energy information systems, realtime systems) can be expressed in terms of international system (SI) metric units or the economic productivity units of time or money. Natural and artificial systems have in common identifiable sets of subsystems which are dedicated to processing information: input transducers, internal transducers, channels and nets, decoders, associators, memory, deciders, encoders, and output transducers. Transducers convert energy into information, or conversely. Intelligent systems can substitute informationintensive behavior for energy-intensive behavior, or for capital- or time-intensive behavior. Ratios of energy used by systems to process information tend to be very small fractions of the total energy used by these systems. However, controversy exists between proponents of the information-communications paradigm and those who regard information as a highly refined form of system control energy. Fresh approaches to information measurement are needed. New generations of real-time systems should provide opportunities to measure relationships between real-time information flows and energy flows in object systems or processes.

Special Issue on Focused Areas and Future Trends of Bio-Inspired Robots “Analysis, Control, and Design for Bio-Inspired Robotics”

Comparing biological and artificial vision systems: Network measures of functional connectivity

Research progress of artificial neural systems based on memristors

Vision is arguably our premier navigational aid, allowing us to map out and actively explore our surroundings. However, we view the world from a constantly shifting platform and some visual mechanisms function optimally only if the images on the retina are reasonably steady. As we go about our everyday activities, visual and vestibular mechanisms help to stabilize our gaze on particular objects of interest by generating eye movements to offset our head movements. The general picture that has emerged of gaze stabilization in primates during motion is of two vestibulo-ocular reflexes, the RVOR and TVOR, that compensate selectively for rotational and translational disturbances of the head, respectively, each with its own independent visual backup mechanisms. A major objective of this chapter is to review recent work on low-level, pre-attentive mechanisms that operate with ultra-short latencies and are largely independent of conscious perception. Recent advances in the field of robotics and, particularly, in the domain of active vision, provide a complementary view of the uses and associated problems of visuo-inertial integration for the stabilization of gaze. Much like biological systems, robots have to comply with physical constraints imposed by the environment and/or by the need to coordinate their sensori-motor components in an efficient way. In contrast with biological systems, however, the experimental variation of implementation parameters and control strategies allows, among other things, a comparison of the different hypotheses and implementations. The goal of this chapter is to draw parallels between the results from biology and from a robot which uses inertial and visual information to stabilize its cameras/eyes. The Chapter is organized as follows. Section 2 describes the peculiarities of the patterns of retinal motion (optic flow) experienced by an observer moving through the environment. The appropriate compensatory eye movements required to stabilize gaze are described in Section 3, introducing the distinction between vergence and version eye movements. Section 4 describes the main characteristics of the vestibular system from a biological and artificial perspective and the distinction between “rotational” and “translational” components of the vestibulo-ocular reflex (VOR). The magnitude of the eye movements required for complete compensation depends on various kinematic parameters such as the position of the eyes in the head, the inter-ocular distance, as well as the distance to the fixation point. Section 5 deals with the integration of visual and inertial information for gaze stabilization. This Section builds upon the concept of “translational” and “rotational” components of the VOR and highlights the differences between “version” and “vergence” control of compensatory eye movements. In particular the role of the radial component of optical flow in the feed-forward control of eye movements is compared with the feed-back loop mediated by binocular disparity. In Section 6, the contribution of inertial and visual information in gaze stabilization is discussed with reference to the different latencies and processing power required by the two modalities. The advantage of the integration of visual and inertial data is discussed from a biological and robotics perspective in the concluding Section.