The energy challenges of artificial superintelligence.

We argue here that contemporary semiconductor computing technology poses a significant if not insurmountable barrier to the emergence of any artificial general intelligence system, let alone one anticipated by many to be "superintelligent". This limit on artificial superintelligence (ASI) emerges from the energy requirements of a system that would be more intelligent but orders of magnitude less efficient in energy use than human brains. An ASI would have to supersede not only a single brain but a large population given the effects of collective behavior on the advancement of societies, further multiplying the energy requirement. A hypothetical ASI would likely consume orders of magnitude more energy than what is available in highly-industrialized nations. We estimate the energy use of ASI with an equation we term the "Erasi equation", for the Energy Requirement for Artificial SuperIntelligence. Additional efficiency consequences will emerge from the current unfocussed and scattered developmental trajectory of AI research. Taken together, these arguments suggest that the emergence of an ASI is highly unlikely in the foreseeable future based on current computer architectures, primarily due to energy constraints, with biomimicry or other new technologies being possible solutions.

Ultraliser: a framework for creating multiscale, high-fidelity and geometrically realistic 3D models for in silico neuroscience.

Ultraliser is a neuroscience-specific software framework capable of creating accurate and biologically realistic 3D models of complex neuroscientific structures at intracellular (e.g. mitochondria and endoplasmic reticula), cellular (e.g. neurons and glia) and even multicellular scales of resolution (e.g. cerebral vasculature and minicolumns). Resulting models are exported as triangulated surface meshes and annotated volumes for multiple applications in in silico neuroscience, allowing scalable supercomputer simulations that can unravel intricate cellular structure-function relationships. Ultraliser implements a high-performance and unconditionally robust voxelization engine adapted to create optimized watertight surface meshes and annotated voxel grids from arbitrary non-watertight triangular soups, digitized morphological skeletons or binary volumetric masks. The framework represents a major leap forward in simulation-based neuroscience, making it possible to employ high-resolution 3D structural models for quantification of surface areas and volumes, which are of the utmost importance for cellular and system simulations. The power of Ultraliser is demonstrated with several use cases in which hundreds of models are created for potential application in diverse types of simulations. Ultraliser is publicly released under the GNU GPL3 license on GitHub (BlueBrain/Ultraliser). SIGNIFICANCE: There is crystal clear evidence on the impact of cell shape on its signaling mechanisms. Structural models can therefore be insightful to realize the function; the more realistic the structure can be, the further we get insights into the function. Creating realistic structural models from existing ones is challenging, particularly when needed for detailed subcellular simulations. We present Ultraliser, a neuroscience-dedicated framework capable of building these structural models with realistic and detailed cellular geometries that can be used for simulations.

Representing stimulus information in an energy metabolism pathway.

We explored a computational model of astrocytic energy metabolism and demonstrated the theoretical plausibility that this type of pathway might be capable of coding information about stimuli in addition to its known functions in cellular energy and carbon budgets. Simulation results indicate that glycogenolytic glycolysis triggered by activation of adrenergic receptors can capture the intensity and duration features of a neuromodulator waveform and can respond in a dose-dependent manner, including non-linear state changes that are analogous to action potentials. We show how this metabolic pathway can translate information about external stimuli to production profiles of energy-carrying molecules such as lactate with a precision beyond simple signal transduction or non-linear amplification. The results suggest the operation of a metabolic state-machine from the spatially discontiguous yet interdependent metabolite elements. Such metabolic pathways might be well-positioned to code an additional level of salient information about a cell's environmental demands to impact its function. Our hypothesis has implications for the computational power and energy efficiency of the brain.

A Standardized Brain Molecular Atlas: A Resource for Systems Modeling and Simulation.

Accurate molecular concentrations are essential for reliable analyses of biochemical networks and the creation of predictive models for molecular and systems biology, yet protein and metabolite concentrations used in such models are often poorly constrained or irreproducible. Challenges of using data from different sources include conflicts in nomenclature and units, as well as discrepancies in experimental procedures, data processing and implementation of the model. To obtain a consistent estimate of protein and metabolite levels, we integrated and normalized data from a large variety of sources to calculate Adjusted Molecular Concentrations. We found a high degree of reproducibility and consistency of many molecular species across brain regions and cell types, consistent with tight homeostatic regulation. We demonstrated the value of this normalization with differential protein expression analyses related to neurodegenerative diseases, brain regions and cell types. We also used the results in proof-of-concept simulations of brain energy metabolism. The standardized Brain Molecular Atlas overcomes the obstacles of missing or inconsistent data to support systems biology research and is provided as a resource for biomolecular modeling.

A Machine-Generated View of the Role of Blood Glucose Levels in the Severity of COVID-19.

SARS-CoV-2 started spreading toward the end of 2019 causing COVID-19, a disease that reached pandemic proportions among the human population within months. The reasons for the spectrum of differences in the severity of the disease across the population, and in particular why the disease affects more severely the aging population and those with specific preconditions are unclear. We developed machine learning models to mine 240,000 scientific articles openly accessible in the CORD-19 database, and constructed knowledge graphs to synthesize the extracted information and navigate the collective knowledge in an attempt to search for a potential common underlying reason for disease severity. The machine-driven framework we developed repeatedly pointed to elevated blood glucose as a key facilitator in the progression of COVID-19. Indeed, when we systematically retraced the steps of the SARS-CoV-2 infection, we found evidence linking elevated glucose to each major step of the life-cycle of the virus, progression of the disease, and presentation of symptoms. Specifically, elevations of glucose provide ideal conditions for the virus to evade and weaken the first level of the immune defense system in the lungs, gain access to deep alveolar cells, bind to the ACE2 receptor and enter the pulmonary cells, accelerate replication of the virus within cells increasing cell death and inducing an pulmonary inflammatory response, which overwhelms an already weakened innate immune system to trigger an avalanche of systemic infections, inflammation and cell damage, a cytokine storm and thrombotic events. We tested the feasibility of the hypothesis by manually reviewing the literature referenced by the machine-generated synthesis, reconstructing atomistically the virus at the surface of the pulmonary airways, and performing quantitative computational modeling of the effects of glucose levels on the infection process. We conclude that elevation in glucose levels can facilitate the progression of the disease through multiple mechanisms and can explain much of the differences in disease severity seen across the population. The study provides diagnostic considerations, new areas of research and potential treatments, and cautions on treatment strategies and critical care conditions that induce elevations in blood glucose levels.

Metaball skinning of synthetic astroglial morphologies into realistic mesh models for visual analytics and in silico simulations.

MOTIVATION: Astrocytes, the most abundant glial cells in the mammalian brain, have an instrumental role in developing neuronal circuits. They contribute to the physical structuring of the brain, modulating synaptic activity and maintaining the blood-brain barrier in addition to other significant aspects that impact brain function. Biophysically, detailed astrocytic models are key to unraveling their functional mechanisms via molecular simulations at microscopic scales. Detailed, and complete, biological reconstructions of astrocytic cells are sparse. Nonetheless, data-driven digital reconstruction of astroglial morphologies that are statistically identical to biological counterparts are becoming available. We use those synthetic morphologies to generate astrocytic meshes with realistic geometries, making it possible to perform these simulations. RESULTS: We present an unconditionally robust method capable of reconstructing high fidelity polygonal meshes of astroglial cells from algorithmically-synthesized morphologies. Our method uses implicit surfaces, or metaballs, to skin the different structural components of astrocytes and then blend them in a seamless fashion. We also provide an end-to-end pipeline to produce optimized two- and three-dimensional meshes for visual analytics and simulations, respectively. The performance of our pipeline has been assessed with a group of 5000 astroglial morphologies and the geometric metrics of the resulting meshes are evaluated. The usability of the meshes is then demonstrated with different use cases. AVAILABILITY AND IMPLEMENTATION: Our metaball skinning algorithm is implemented in Blender 2.82 relying on its Python API (Application Programming Interface). To make it accessible to computational biologists and neuroscientists, the implementation has been integrated into NeuroMorphoVis, an open source and domain specific package that is primarily designed for neuronal morphology visualization and meshing. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Interactive visualization and analysis of morphological skeletons of brain vasculature networks with VessMorphoVis.

MOTIVATION: Accurate morphological models of brain vasculature are key to modeling and simulating cerebral blood flow in realistic vascular networks. This in silico approach is fundamental to revealing the principles of neurovascular coupling. Validating those vascular morphologies entails performing certain visual analysis tasks that cannot be accomplished with generic visualization frameworks. This limitation has a substantial impact on the accuracy of the vascular models employed in the simulation. RESULTS: We present VessMorphoVis, an integrated suite of toolboxes for interactive visualization and analysis of vast brain vascular networks represented by morphological graphs segmented originally from imaging or microscopy stacks. Our workflow leverages the outstanding potentials of Blender, aiming to establish an integrated, extensible and domain-specific framework capable of interactive visualization, analysis, repair, high-fidelity meshing and high-quality rendering of vascular morphologies. Based on the initial feedback of the users, we anticipate that our framework will be an essential component in vascular modeling and simulation in the future, filling a gap that is at present largely unfulfilled. AVAILABILITY AND IMPLEMENTATION: VessMorphoVis is freely available under the GNU public license on Github at https://github.com/BlueBrain/VessMorphoVis. The morphology analysis, visualization, meshing and rendering modules are implemented as an add-on for Blender 2.8 based on its Python API (application programming interface). The add-on functionality is made available to users through an intuitive graphical user interface, as well as through exhaustive configuration files calling the API via a feature-rich command line interface running Blender in background mode. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Excitation states of metabolic networks predict dose-response fingerprinting and ligand pulse phase signalling.

With a computational model of energy metabolism in an astrocyte, we show how a system of enzymes in a cascade can act as a functional unit of interdependent reactions, rather than merely a series of independent reactions. These systems may exist in multiple states, depending on the level of stimulation, and the effects of substrates at any point will depend on those states. Response trajectories of metabolites downstream from cAMP-stimulated glycogenolysis exhibit a host of non-linear dynamical response characteristics including hysteresis and response envelopes. Dose-dependent phase transitions predict a novel intracellular signalling mechanism and suggest a theoretical framework that could be relevant to single cell information processing, drug discovery or synthetic biology. Ligands may produce unique dose-response fingerprints depending on the state of the system, allowing selective output tuning. We conclude with the observation that state- and dose-dependent phase transitions, what we dub "ligand pulses" (LPs), may carry information and resemble action potentials (APs) generated from excitatory postsynaptic potentials. In our model, the relevant information from a cAMP-dependent glycolytic cascade in astrocytes could reflect the level of neuromodulatory input that signals an energy demand threshold. We propose that both APs and LPs represent specialized cases of molecular phase signalling with a common evolutionary root.

A Process for Digitizing and Simulating Biologically Realistic Oligocellular Networks Demonstrated for the Neuro-Glio-Vascular Ensemble.

One will not understand the brain without an integrated exploration of structure and function, these attributes being two sides of the same coin: together they form the currency of biological computation. Accordingly, biologically realistic models require the re-creation of the architecture of the cellular components in which biochemical reactions are contained. We describe here a process of reconstructing a functional oligocellular assembly that is responsible for energy supply management in the brain and creating a computational model of the associated biochemical and biophysical processes. The reactions that underwrite thought are both constrained by and take advantage of brain morphologies pertaining to neurons, astrocytes and the blood vessels that deliver oxygen, glucose and other nutrients. Each component of this neuro-glio-vasculature ensemble (NGV) carries-out delegated tasks, as the dynamics of this system provide for each cell-type its own energy requirements while including mechanisms that allow cooperative energy transfers. Our process for recreating the ultrastructure of cellular components and modeling the reactions that describe energy flow uses an amalgam of state-of the-art techniques, including digital reconstructions of electron micrographs, advanced data analysis tools, computational simulations and in silico visualization software. While we demonstrate this process with the NGV, it is equally well adapted to any cellular system for integrating multimodal cellular data in a coherent framework.

Norepinephrine stimulates glycogenolysis in astrocytes to fuel neurons with lactate.

The mechanism of rapid energy supply to the brain, especially to accommodate the heightened metabolic activity of excited states, is not well-understood. We explored the role of glycogen as a fuel source for neuromodulation using the noradrenergic stimulation of glia in a computational model of the neural-glial-vasculature ensemble (NGV). The detection of norepinephrine (NE) by the astrocyte and the coupled cAMP signal are rapid and largely insensitive to the distance of the locus coeruleus projection release sites from the glia, implying a diminished impact for volume transmission in high affinity receptor transduction systems. Glucosyl-conjugated units liberated from glial glycogen by NE-elicited cAMP second messenger transduction winds sequentially through the glycolytic cascade, generating robust increases in NADH and ATP before pyruvate is finally transformed into lactate. This astrocytic lactate is rapidly exported by monocarboxylate transporters to the associated neuron, demonstrating that the astrocyte-to-neuron lactate shuttle activated by glycogenolysis is a likely fuel source for neuromodulation and enhanced neural activity. Altogether, the energy supply for both astrocytes and neurons can be supplied rapidly by glycogenolysis upon neuromodulatory stimulus.

Physiological Dynamics in Demyelinating Diseases: Unraveling Complex Relationships through Computer Modeling.

Despite intense research, few treatments are available for most neurological disorders. Demyelinating diseases are no exception. This is perhaps not surprising considering the multifactorial nature of these diseases, which involve complex interactions between immune system cells, glia and neurons. In the case of multiple sclerosis, for example, there is no unanimity among researchers about the cause or even which system or cell type could be ground zero. This situation precludes the development and strategic application of mechanism-based therapies. We will discuss how computational modeling applied to questions at different biological levels can help link together disparate observations and decipher complex mechanisms whose solutions are not amenable to simple reductionism. By making testable predictions and revealing critical gaps in existing knowledge, such models can help direct research and will provide a rigorous framework in which to integrate new data as they are collected. Nowadays, there is no shortage of data; the challenge is to make sense of it all. In that respect, computational modeling is an invaluable tool that could, ultimately, transform how we understand, diagnose, and treat demyelinating diseases.

Cooperativity between remote sites of ectopic spiking allows afterdischarge to be initiated and maintained at different locations.

Many symptoms of nerve damage arise from ectopic spiking caused by hyperexcitability. Ectopic spiking can originate at the site of axonal damage and elsewhere within affected neurons. This raises the question of whether localized damage elicits cell-wide changes in excitability and/or if localized changes in excitability can drive abnormal spiking at remote locations. Computer modeling revealed an example of the latter involving afterdischarge (AD)--stimulus-evoked spiking that outlasts stimulation. We found that AD originating in a hyperexcitable region of axon could shift to the soma where it was maintained. This repositioning of ectopic spike initiation was independent of distance between the two sites but relied on the rate and number of ectopic spikes originating from the first site. Nonlinear dynamical analysis of a reduced model demonstrated that properties which rendered the axonal site prone to initiating AD discouraged it from maintaining AD, whereas the soma had the inverse properties thus enabling the two sites to interact cooperatively. A first phase of AD originating in the axon could, by providing sufficient drive to trigger somatic AD, give way to a second phase of AD originating in the soma such that spiking continued when axonal AD failed. Ectopic spikes originating from the soma during phase 2 AD propagated successfully through the defunct site of axonal spike initiation. This novel mechanism whereby ectopic spiking at one site facilitates ectopic spiking at another site is likely to contribute to the chronification of hyperexcitability in conditions such as neuropathic pain.

Multi-timescale modeling of activity-dependent metabolic coupling in the neuron-glia-vasculature ensemble.

Glucose is the main energy substrate in the adult brain under normal conditions. Accumulating evidence, however, indicates that lactate produced in astrocytes (a type of glial cell) can also fuel neuronal activity. The quantitative aspects of this so-called astrocyte-neuron lactate shuttle (ANLS) are still debated. To address this question, we developed a detailed biophysical model of the brain's metabolic interactions. Our model integrates three modeling approaches, the Buxton-Wang model of vascular dynamics, the Hodgkin-Huxley formulation of neuronal membrane excitability and a biophysical model of metabolic pathways. This approach provides a template for large-scale simulations of the neuron-glia-vasculature (NGV) ensemble, and for the first time integrates the respective timescales at which energy metabolism and neuronal excitability occur. The model is constrained by relative neuronal and astrocytic oxygen and glucose utilization, by the concentration of metabolites at rest and by the temporal dynamics of NADH upon activation. These constraints produced four observations. First, a transfer of lactate from astrocytes to neurons emerged in response to activity. Second, constrained by activity-dependent NADH transients, neuronal oxidative metabolism increased first upon activation with a subsequent delayed astrocytic glycolysis increase. Third, the model correctly predicted the dynamics of extracellular lactate and oxygen as observed in vivo in rats. Fourth, the model correctly predicted the temporal dynamics of tissue lactate, of tissue glucose and oxygen consumption, and of the BOLD signal as reported in human studies. These findings not only support the ANLS hypothesis but also provide a quantitative mathematical description of the metabolic activation in neurons and glial cells, as well as of the macroscopic measurements obtained during brain imaging.

Proposed evolutionary changes in the role of myelin.

Myelin is the multi-layered lipid sheet periodically wrapped around neuronal axons. It is most frequently found in vertebrates. Myelin allows for saltatory action potential (AP) conduction along axons. During this form of conduction, the AP travels passively along the myelin-covered part of the axon, and is recharged at the intermittent nodes of Ranvier. Thus, myelin can reduce the energy load needed and/or increase the speed of AP conduction. Myelin first evolved during the Ordovician period. We hypothesize that myelin's first role was mainly energy conservation. During the later "Mesozoic marine revolution," marine ecosystems changed toward an increase in marine predation pressure. We hypothesize that the main purpose of myelin changed from energy conservation to conduction speed increase during this Mesozoic marine revolution. To test this hypothesis, we optimized models of myelinated axons for a combination of AP conduction velocity and energy efficiency. We demonstrate that there is a trade-off between these objectives. We then compared the simulation results to empirical data and conclude that while the data are consistent with the theory, additional measurements are necessary for a complete evaluation of the proposed hypothesis.

Explaining pathological changes in axonal excitability through dynamical analysis of conductance-based models.

Neurons rely on action potentials, or spikes, to relay information. Pathological changes in spike generation likely contribute to certain enigmatic features of neurological disease, like paroxysmal attacks of pain and muscle spasm. Paroxysmal symptoms are characterized by abrupt onset and short duration, and are associated with abnormal spiking although the exact pathophysiology remains unclear. To help decipher the biophysical basis for 'paroxysmal' spiking, we replicated afterdischarge (i.e. continued spiking after a brief stimulus) in a minimal conductance-based axon model. We then applied nonlinear dynamical analysis to explain the dynamical basis for initiation and termination of afterdischarge. A perturbation could abruptly switch the system between two (quasi-)stable attractor states: rest and repetitive spiking. This bistability was a consequence of slow positive feedback mediated by persistent inward current. Initiation of afterdischarge was explained by activation of the persistent inward current forcing the system to cross a saddle point that separates the basins of attraction associated with each attractor. Termination of afterdischarge was explained by the attractor associated with repetitive spiking being destroyed. This occurred when ultra-slow negative feedback, such as intracellular sodium accumulation, caused the saddle point and stable limit cycle to collide; in that regard, the active attractor is not truly stable when the slowest dynamics are taken into account. The model also explains other features of paroxysmal symptoms, including temporal summation and refractoriness.

Imbalance of ionic conductances contributes to diverse symptoms of demyelination.

Fast axonal conduction of action potentials in mammals relies on myelin insulation. Demyelination can cause slowed, blocked, desynchronized, or paradoxically excessive spiking that underlies the symptoms observed in demyelination diseases. The diversity and timing of such symptoms are poorly understood, often intermittent, and uncorrelated with disease progress. We modeled the effects of demyelination (and secondary remodeling) on intrinsic axonal excitability using Hodgkin-Huxley and reduced Morris-Lecar models. Simulations and analysis suggested a simple explanation for the breadth of symptoms and revealed that the ratio of sodium to leak conductance, g(Na)/g(L), acted as a four-way switch controlling excitability patterns that included spike failure, single spike transmission, afterdischarge, and spontaneous spiking. Failure occurred when this ratio fell below a threshold value. Afterdischarge occurred at g(Na)/g(L) just below the threshold for spontaneous spiking and required a slow inward current that allowed for two stable attractor states, one corresponding to quiescence and the other to repetitive spiking. A neuron prone to afterdischarge could function normally unless it was switched to its "pathological" attractor state; thus, although the underlying pathology may develop slowly by continuous changes in membrane conductances, a discontinuous change in axonal excitability can occur and lead to paroxysmal symptoms. We conclude that tonic and paroxysmal positive symptoms as well as negative symptoms may be a consequence of varying degrees of imbalance between g(Na) and g(L) after demyelination. The KCNK family of g(L) potassium channels may be an important target for new drugs to treat the symptoms of demyelination.

Rapid and modifiable neurotransmitter receptor dynamics at a neuronal synapse in vivo.

Synaptic plasticity underlies the adaptability of the mammalian brain, but has been difficult to study in living animals. Here we imaged the synapses between pre- and postganglionic neurons in the mouse submandibular ganglion in vivo, focusing on the mechanisms that maintain and regulate neurotransmitter receptor density at postsynaptic sites. Normally, synaptic receptor densities were maintained by rapid exchange of receptors with nonsynaptic regions (over minutes) and by continual turnover of cell surface receptors (over hours). However, after ganglion cell axons were crushed, synaptic receptors showed greater lateral mobility and there was a precipitous decline in insertion. These changes led to near-complete loss of synaptic receptors and synaptic depression. Disappearance of postsynaptic spines and presynaptic terminals followed this acute synaptic depression. Therefore, neurotransmitter receptor dynamism associated with rapid changes in synaptic efficacy precedes long-lasting structural changes in synaptic connectivity.

Computational modeling of three-dimensional electrodiffusion in biological systems: application to the node of Ranvier.

A computational model is presented for the simulation of three-dimensional electrodiffusion of ions. Finite volume techniques were used to solve the Poisson-Nernst-Planck equation, and a dual Delaunay-Voronoi mesh was constructed to evaluate fluxes of ions, as well as resulting electric potentials. The algorithm has been validated and applied to a generalized node of Ranvier, where numerical results for computed action potentials agree well with cable model predictions for large clusters of voltage-gated ion channels. At smaller channel clusters, however, the three-dimensional electrodiffusion predictions diverge from the cable model predictions and show a broadening of the action potential, indicating a significant effect due to each channel's own local electric field. The node of Ranvier complex is an elaborate organization of membrane-bound aqueous compartments, and the model presented here represents what we believe is a significant first step in simulating electrophysiological events with combined realistic structural and physiological data.

Evidence for ectopic neurotransmission at a neuronal synapse.

Neurotransmitter release is well known to occur at specialized synaptic regions that include presynaptic active zones and postsynaptic densities. At cholinergic synapses in the chick ciliary ganglion, however, membrane formations and physiological measurements suggest that release distant from postsynaptic densities can activate the predominantly extrasynaptic alpha7 nicotinic receptor subtype. We explored such ectopic neurotransmission with a novel model synapse that combines Monte Carlo simulations with high-resolution serial electron microscopic tomography. Simulated synaptic activity is consistent with experimental recordings of miniature excitatory postsynaptic currents only when ectopic transmission is included in the model, broadening the possibilities for mechanisms of neuronal communication.

Age-associated synapse elimination in mouse parasympathetic ganglia.

Little is known about the effects of aging on synapses in the mammalian nervous system. We examined the innervation of individual mouse submandibular ganglion (SMG) neurons for evidence of age-related changes in synapse efficacy and number. For approximately 85% of adult life expectancy (30 months) the efficacy of synaptic transmission, as determined by excitatory postsynaptic potential (EPSP) amplitudes, remains constant. Similarly, the number of synapses contacting individual SMG neurons is also unchanged. After 30 months of age, however, some neurons (23%) dramatically lose synaptic input exhibiting both smaller EPSP amplitude and fewer synaptic boutons. Attenuation of both the amplitude and frequency of miniature EPSPs was also observed in neurons from aged animals. Electron micrographs revealed that, although there were many vesicle-laden preganglionic axonal processes in the vicinity of the postsynaptic membrane, the number of synaptic contacts was significantly lower in old animals. These results demonstrate primary, age-associated synapse elimination with functional consequences that cannot be explained by pre- or postsynaptic cell death.