Recent advances in deep learning have allowed artificial intelligence (AI) to reach near human-level performance in many sensory, perceptual, linguistic, and cognitive tasks. There is a growing need, ...however, for novel, brain-inspired cognitive architectures. The Global Workspace Theory (GWT) refers to a large-scale system integrating and distributing information among networks of specialized modules to create higher-level forms of cognition and awareness. We argue that the time is ripe to consider explicit implementations of this theory using deep-learning techniques. We propose a roadmap based on unsupervised neural translation between multiple latent spaces (neural networks trained for distinct tasks, on distinct sensory inputs and/or modalities) to create a unique, amodal Global Latent Workspace (GLW). Potential functional advantages of GLW are reviewed, along with neuroscientific implications.
In recent years, deep learning has steadily improved the state of the art in artificial intelligence (AI), but mainly for single, well-defined tasks or challenges.Novel advanced neural network architectures, possibly inspired by neuroscience, are needed to create more general-purpose AI systems with flexible and robust capabilities.The Global Workspace Theory (GWT), introduced over 30 years ago, proposed such an architecture; we now consider its implementation in a deep-learning framework.The global workspace was primarily designed to account for conscious information processing in the human brain, but in principle, its associated functional advantages could generalize to artificial systems.In turn, considering an artificial global workspace can help constrain neuroscientific investigations of brain function and consciousness.
This paper considers neuronal architectures from a computational perspective and asks what aspects of neuroanatomy and neurophysiology can be disclosed by the nature of neuronal computations? In ...particular, we extend current formulations of the brain as an organ of inference—based upon hierarchical predictive coding—and consider how these inferences are orchestrated. In other words, what would the brain require to dynamically coordinate and contextualize its message passing to optimize its computational goals? The answer that emerges rests on the delicate (modulatory) gain control of neuronal populations that select and coordinate (prediction error) signals that ascend cortical hierarchies. This is important because it speaks to a hierarchical anatomy of extrinsic (between region) connections that form two distinct classes, namely a class of driving (first-order) connections that are concerned with encoding the content of neuronal representations and a class of modulatory (second-order) connections that establish context—in the form of the salience or precision ascribed to content. We explore the implications of this distinction from a formal perspective (using simulations of feature–ground segregation) and consider the neurobiological substrates of the ensuing precision-engineered dynamics, with a special focus on the pulvinar and attention.
When visual input has conflicting interpretations, conscious perception can alternate spontaneously between competing interpretations 1. There is a large amount of unexplained variability between ...individuals in the rate of such spontaneous alternations in perception 2–5. We hypothesized that variability in perceptual rivalry might be reflected in individual differences in brain structure, because brain structure can exhibit systematic relationships with an individual's cognitive experiences and skills 6–9. To test this notion, we examined in a large group of individuals how cortical thickness, local gray-matter density, and local white-matter integrity correlate with individuals' alternation rate for a bistable, rotating structure-from-motion stimulus 10. All of these macroscopic measures of brain structure consistently revealed that the structure of bilateral superior parietal lobes (SPL) could account for interindividual variability in perceptual alternation rate. Furthermore, we examined whether the bilateral SPL regions play a causal role in the rate of perceptual alternations by using transcranial magnetic stimulation (TMS) and found that transient disruption of these areas indeed decreases the rate of perceptual alternations. These findings demonstrate a direct relationship between structure of SPL and individuals' perceptual switch rate.
Display omitted
► Structure of superior parietal lobe (SPL) predicts switch rate in perceptual rivalry ► White-matter integrity in SPL correlates with individuals' switch rate ► Deactivation of SPL with transcranial magnetic stimulation slows perceptual rivalry
We summarize the original formulation of the free energy principle and highlight some technical issues. We discuss how these issues affect related results involving generalised coordinates and, where ...appropriate, mention consequences for and reveal, up to now unacknowledged, differences from newer formulations of the free energy principle. In particular, we reveal that various definitions of the "Markov blanket" proposed in different works are not equivalent. We show that crucial steps in the free energy argument, which involve rewriting the equations of motion of systems with Markov blankets, are not generally correct without additional (previously unstated) assumptions. We prove by counterexamples that the original free energy lemma, when taken at face value, is wrong. We show further that this free energy lemma, when it does hold, implies the equality of variational density and ergodic conditional density. The interpretation in terms of Bayesian inference hinges on this point, and we hence conclude that it is not sufficiently justified. Additionally, we highlight that the variational densities presented in newer formulations of the free energy principle and lemma are parametrised by different variables than in older works, leading to a substantially different interpretation of the theory. Note that we only highlight some specific problems in the discussed publications. These problems do not rule out conclusively that the general ideas behind the free energy principle are worth pursuing.
Media multitasking, or the concurrent consumption of multiple media forms, is increasingly prevalent in today's society and has been associated with negative psychosocial and cognitive impacts. ...Individuals who engage in heavier media-multitasking are found to perform worse on cognitive control tasks and exhibit more socio-emotional difficulties. However, the neural processes associated with media multi-tasking remain unexplored. The present study investigated relationships between media multitasking activity and brain structure. Research has demonstrated that brain structure can be altered upon prolonged exposure to novel environments and experience. Thus, we expected differential engagements in media multitasking to correlate with brain structure variability. This was confirmed via Voxel-Based Morphometry (VBM) analyses: Individuals with higher Media Multitasking Index (MMI) scores had smaller gray matter density in the anterior cingulate cortex (ACC). Functional connectivity between this ACC region and the precuneus was negatively associated with MMI. Our findings suggest a possible structural correlate for the observed decreased cognitive control performance and socio-emotional regulation in heavy media-multitaskers. While the cross-sectional nature of our study does not allow us to specify the direction of causality, our results brought to light novel associations between individual media multitasking behaviors and ACC structure differences.
Celotno besedilo
Dostopno za:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
Around 20% of the population exhibits moderate to severe numerical disabilities 1–3, and a further percentage loses its numerical competence during the lifespan as a result of stroke or degenerative ...diseases 4. In this work, we investigated the feasibility of using noninvasive stimulation to the parietal lobe during numerical learning to selectively improve numerical abilities. We used transcranial direct current stimulation (TDCS), a method that can selectively inhibit or excitate neuronal populations by modulating GABAergic (anodal stimulation) and glutamatergic (cathodal stimulation) activity 5, 6. We trained subjects for 6 days with artificial numerical symbols, during which we applied concurrent TDCS to the parietal lobes. The polarity of the brain stimulation specifically enhanced or impaired the acquisition of automatic number processing and the mapping of number into space, both important indices of numerical proficiency 7–9. The improvement was still present 6 months after the training. Control tasks revealed that the effect of brain stimulation was specific to the representation of artificial numerical symbols. The specificity and longevity of TDCS on numerical abilities establishes TDCS as a realistic tool for intervention in cases of atypical numerical development or loss of numerical abilities because of stroke or degenerative illnesses.
► Brain stimulation to the parietal cortex can enhance or impair numerical abilities ► The effects were specific to the polarity of the current ► The improvement in numerical abilities lasts up to 6 months ► The brain stimulation affected specifically the material that was recently learned
Individual differences in the structure of parietal and prefrontal cortex predict the stability of bistable visual perception. However, the mechanisms linking such individual differences in brain ...structures to behaviour remain elusive. Here we demonstrate a systematic relationship between the dynamics of brain activity, cortical structure and behaviour underpinning bistable perception. Using fMRI in humans, we find that the activity dynamics during bistable perception are well described as fluctuating between three spatially distributed energy minimums: visual-area-dominant, frontal-area-dominant and intermediate states. Transitions between these energy minimums predicted behaviour, with participants whose brain activity tend to reflect the visual-area-dominant state exhibiting more stable perception and those whose activity transits to frontal-area-dominant states reporting more frequent perceptual switches. Critically, these brain activity dynamics are correlated with individual differences in grey matter volume of the corresponding brain areas. Thus, individual differences in the large-scale dynamics of brain activity link focal brain structure with bistable perception.
Behavioral and neuroscientific studies explore two pathways through which internalized social norms promote prosocial behavior. One pathway involves internal control of impulsive selfishness, and the ...other involves emotion-based prosocial preferences that are translated into behavior when they evade cognitive control for pursuing self-interest.Wemeasured 443 participants’ overall prosocial behavior in four economic games. Participants’ predispositions social value orientation (SVO) were more strongly reflected in their overall game behavior when they made decisions quickly than when they spent a longer time. Prosocially (or selfishly) predisposed participants behaved less prosocially (or less selfishly) when they spent more time in decision making, such that their SVO prosociality yielded limited effects in actual behavior in their slow decisions. The increase (or decrease) in slower decision makers was prominent among consistent prosocials (or proselfs) whose strong preference for prosocial (or proself) goals would make it less likely to experience conflict between prosocial and proself goals. The strong effect of RT on behavior in consistent prosocials (or proselfs) suggests that conflict between prosocial and selfish goals alone is not responsible for slow decisions. Specifically, we found that contemplation of the risk of being exploited by others (social risk aversion) was partly responsible for making consistent prosocials (but not consistent proselfs) spend longer time in decision making and behave less prosocially. Conflict between means rather than between goals (immediate versus strategic pursuit of self-interest) was suggested to be responsible for the time-related increase in consistent proselfs’ prosocial behavior. The findings of this study are generally in favor of the intuitive cooperation model of prosocial behavior.
Professional ball game players report the feeling of the ball ‘slowing-down’ before hitting it. Because effective motor preparation is critical in achieving such expert motor performance, these ...anecdotal comments imply that the subjective passage of time may be influenced by preparation for action. Previous reports of temporal illusions associated with action generally emphasize compensation for suppressed sensory signals that accompany motor commands. Here, we show that the time is perceived slowed-down during preparation of a ballistic reaching movement before action, involving enhancement of sensory processing. Preparing for a reaching movement increased perceived duration of a visual stimulus. This effect was tightly linked to action preparation, because the amount of temporal dilation increased with the information about the upcoming movement. Furthermore, we showed a reduction of perceived frequency for flickering stimuli and an enhanced detection of rapidly presented letters during action preparation, suggesting increased temporal resolution of visual perception during action preparation. We propose that the temporal dilation during action preparation reflects the function of the brain to maximize the capacity of sensory information-acquisition prior to execution of a ballistic movement. This strategy might facilitate changing or inhibiting the planned action in response to last-minute changes in the external environment.
The neural mechanisms underlying conscious visual perception have been extensively investigated using bistable perception paradigms. Previous functional magnetic resonance imaging (fMRI) and ...transcranial magnetic stimulation (TMS) studies suggest that the right anterior superior parietal (r-aSPL) and the right posterior superior parietal lobule (r-pSPL) have opposite roles in triggering perceptual reversals. It has been proposed that these two areas are part of a hierarchical network whose dynamics determine perceptual switches. However, how these two parietal regions interact with each other and with the rest of the brain during bistable perception is not known. Here, we investigated such a model by recording brain activity using fMRI while participants viewed a bistable structure-from-motion stimulus. Using dynamic causal modeling (DCM), we found that resolving such perceptual ambiguity was specifically associated with reciprocal interactions between these parietal regions and V5/MT. Strikingly, the strength of bottom-up coupling between V5/MT to r-pSPL and from r-pSPL to r-aSPL predicted individual mean dominance duration. Our findings are consistent with a hierarchical predictive coding model of parietal involvement in bistable perception and suggest that visual information processing underlying spontaneous perceptual switches can be described as changes in connectivity strength between parietal and visual cortical regions.
PDF
