{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,1,23]],"date-time":"2026-01-23T14:12:45Z","timestamp":1769177565045,"version":"3.49.0"},"reference-count":85,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2024,6,14]],"date-time":"2024-06-14T00:00:00Z","timestamp":1718323200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Comput. Neurosci."],"abstract":"<jats:p>As the apparent intelligence of artificial neural networks (ANNs) advances, they are increasingly likened to the functional networks and information processing capabilities of the human brain. Such comparisons have typically focused on particular modalities, such as vision or language. The next frontier is to use the latest advances in ANNs to design and investigate scalable models of higher-level cognitive processes, such as conscious information access, which have historically lacked concrete and specific hypotheses for scientific evaluation. In this work, we propose and then empirically assess an embodied agent with a structure based on global workspace theory (GWT) as specified in the recently proposed \u201cindicator properties\u201d of consciousness. In contrast to prior works on GWT which utilized single modalities, our agent is trained to navigate 3D environments based on realistic audiovisual inputs. We find that the global workspace architecture performs better and more robustly at smaller working memory sizes, as compared to a standard recurrent architecture. Beyond performance, we perform a series of analyses on the learned representations of our architecture and share findings that point to task complexity and regularization being essential for feature learning and the development of meaningful attentional patterns within the workspace.<\/jats:p>","DOI":"10.3389\/fncom.2024.1352685","type":"journal-article","created":{"date-parts":[[2024,6,14]],"date-time":"2024-06-14T05:17:21Z","timestamp":1718342241000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":4,"title":["Design and evaluation of a global workspace agent embodied in a realistic multimodal environment"],"prefix":"10.3389","volume":"18","author":[{"given":"Rousslan Fernand Julien","family":"Dossa","sequence":"first","affiliation":[]},{"given":"Kai","family":"Arulkumaran","sequence":"additional","affiliation":[]},{"given":"Arthur","family":"Juliani","sequence":"additional","affiliation":[]},{"given":"Shuntaro","family":"Sasai","sequence":"additional","affiliation":[]},{"given":"Ryota","family":"Kanai","sequence":"additional","affiliation":[]}],"member":"1965","published-online":{"date-parts":[[2024,6,14]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"99","DOI":"10.1101\/sqb.2014.79.024729","article-title":"Neural mechanisms underlying visual object recognition","volume":"79","author":"Afraz","year":"2014","journal-title":"Cold Spring Harb. Symp. Quant. Biol"},{"key":"B2","first-page":"29304","article-title":"Deep reinforcement learning at the edge of the statistical precipice","volume":"34","author":"Agarwal","year":"2021","journal-title":"Adv. Neural Inf. Process. Syst"},{"key":"B3","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1910.07113","article-title":"Solving Rubik's cube with a robot hand","author":"Akkaya","year":"2019","journal-title":"arXiv"},{"key":"B4","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1610.01644","article-title":"Understanding intermediate layers using linear classifier probes","author":"Alain","year":"2016","journal-title":"arXiv"},{"key":"B5","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv:1607.06450","article-title":"Layer normalization","author":"Ba","year":"2016","journal-title":"arXiv"},{"key":"B6","volume-title":"A Cognitive Theory of Consciousness","author":"Baars","year":"1993"},{"key":"B7","doi-asserted-by":"publisher","first-page":"45","DOI":"10.1016\/S0079-6123(05)50004-9","article-title":"Global workspace theory of consciousness: toward a cognitive neuroscience of human experience","volume":"150","author":"Baars","year":"2005","journal-title":"Prog. Brain Res"},{"key":"B8","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv:1409.0473","article-title":"Neural machine translation by jointly learning to align and translate","author":"Bahdanau","year":"2014","journal-title":"arXiv"},{"key":"B9","volume-title":"Emergent Tool use from Multi-Agent Interaction","author":"Baker","year":"2019"},{"key":"B10","first-page":"29304","article-title":"\u201cBio-inspired memory generation by recurrent neural networks,\u201d","volume-title":"International Work-Conference on Artificial Neural Networks, Volume 34","author":"Bedia","year":"2007"},{"key":"B11","doi-asserted-by":"publisher","first-page":"49","DOI":"10.1162\/tacl_a_00254","article-title":"Analysis methods in neural language processing: a survey","volume":"7","author":"Belinkov","year":"2019","journal-title":"Trans. Assoc. Comput. Linguist"},{"key":"B12","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1709.08568","article-title":"The consciousness prior","author":"Bengio","year":"2017","journal-title":"arXiv"},{"key":"B13","unstructured":"BiewaldL.\n          Experiment Tracking With Weights and Biases2020"},{"key":"B14","doi-asserted-by":"publisher","first-page":"227","DOI":"10.1017\/S0140525X00038188","article-title":"On a confusion about a function of consciousness","volume":"18","author":"Block","year":"1995","journal-title":"Behav. Brain Sci"},{"key":"B15","doi-asserted-by":"publisher","first-page":"e2115934119","DOI":"10.1073\/pnas.2115934119","article-title":"A theory of consciousness from a theoretical computer science perspective: insights from the conscious turing machine","volume":"119","author":"Blum","year":"2022","journal-title":"Proc. Nat. Acad. Sci"},{"key":"B16","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv:2308.08708","article-title":"Consciousness in artificial intelligence: insights from the science of consciousness","author":"Butlin","year":"2023","journal-title":"arXiv"},{"key":"B17","doi-asserted-by":"publisher","DOI":"10.1101\/2020.07.03.186288","article-title":"Language processing in brains and deep neural networks: computational convergence and its limits","author":"Caucheteux","year":"2020","journal-title":"bioRxiv"},{"key":"B18","doi-asserted-by":"publisher","first-page":"134","DOI":"10.1038\/s42003-022-03036-1","article-title":"Brains and algorithms partially converge in natural language processing","volume":"5","author":"Caucheteux","year":"2022","journal-title":"Commun. Biol"},{"key":"B19","first-page":"200","article-title":"Facing up to the problem of consciousness","volume":"2","author":"Chalmers","year":"1995","journal-title":"J. Conscious. Stud"},{"key":"B20","first-page":"15516","article-title":"\u201cSemantic audio-visual navigation,\u201d","volume-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition","author":"Chen","year":"2021"},{"key":"B21","first-page":"17","article-title":"\u201cSoundspaces: audio-visual navigation in 3D environments,\u201d","volume-title":"ECCV","author":"Chen","year":"2020"},{"key":"B22","first-page":"8896","article-title":"Soundspaces 2.0: a simulation platform for visual-acoustic learning","volume":"35","author":"Chen","year":"2022","journal-title":"Adv. Neural Inf. Process. Syst"},{"key":"B23","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1409.1259","article-title":"On the properties of neural machine translation: encoder-decoder approaches","author":"Cho","year":"2014","journal-title":"arXiv"},{"key":"B24","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1906.04341","article-title":"What does bert look at? An analysis of BERT's attention","author":"Clark","year":"2019","journal-title":"arXiv"},{"key":"B25","first-page":"2048","article-title":"\u201cLeveraging procedural generation to benchmark reinforcement learning,\u201d","volume-title":"Proceedings of the 37th International Conference on Machine Learning","author":"Cobbe","year":"2020"},{"key":"B26","doi-asserted-by":"publisher","first-page":"143","DOI":"10.1016\/j.neucom.2022.04.005","article-title":"Analysing deep reinforcement learning agents trained with domain randomisation","volume":"493","author":"Dai","year":"2022","journal-title":"Neurocomputing"},{"key":"B27","doi-asserted-by":"crossref","first-page":"43","DOI":"10.1007\/978-3-030-54173-6_4","article-title":"\u201cWhat is consciousness, and could machines have it?\u201d","volume-title":"Robotics, AI, and Humanity: Science, Ethics, and Policy","author":"Dehaene","year":"2021"},{"key":"B28","doi-asserted-by":"publisher","first-page":"20170342","DOI":"10.1098\/rstb.2017.0342","article-title":"Facing up to the hard question of consciousness","volume":"373","author":"Dennett","year":"2018","journal-title":"Philos. Trans. R. Soc. B: Biol. Sci"},{"key":"B29","unstructured":"DhariwalP.\n            HesseC.\n            KlimovO.\n            NicholA.\n            PlappertM.\n            RadfordA.\n          OpenAI Baselines2017"},{"key":"B30","doi-asserted-by":"publisher","first-page":"827","DOI":"10.1038\/s42003-021-02341-5","article-title":"A convolutional neural-network framework for modelling auditory sensory cells and synapses","volume":"4","author":"Drakopoulos","year":"2021","journal-title":"Commun. Biol"},{"key":"B31","first-page":"1407","article-title":"\u201cImpala: scalable distributed deep-Rl with importance weighted actor-learner architectures,\u201d","volume-title":"Proceedings of the 35th International Conference on Machine Learning","author":"Espeholt","year":"2018"},{"key":"B32","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv:2103.06257","article-title":"Maximum entropy RL (Provably) solves some robust RL problems","author":"Eysenbach","year":"2021","journal-title":"arXiv"},{"key":"B33","doi-asserted-by":"publisher","first-page":"499","DOI":"10.1080\/019697297126029","article-title":"Autonomous agents as embodied AI","volume":"28","author":"Franklin","year":"1997","journal-title":"Cybern. Syst"},{"key":"B34","doi-asserted-by":"publisher","first-page":"121","DOI":"10.1007\/BF00342633","article-title":"Cognitron: a self-organizing multilayered neural network","volume":"20","author":"Fukushima","year":"1975","journal-title":"Biol. Cybern"},{"key":"B35","doi-asserted-by":"publisher","first-page":"193","DOI":"10.1007\/BF00344251","article-title":"Neocognitron: a self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position","volume":"36","author":"Fukushima","year":"1980","journal-title":"Biol. Cybern"},{"key":"B36","doi-asserted-by":"crossref","DOI":"10.1017\/CBO9780511805844","volume-title":"Embodiment and Cognitive Sciences","author":"Gibbs Jr","year":"2005"},{"key":"B37","volume-title":"Deep Learning","author":"Goodfellow","year":"2016"},{"key":"B38","doi-asserted-by":"publisher","first-page":"20210068","DOI":"10.1098\/rspa.2021.0068","article-title":"Inductive Biases for deep learning of higher-level cognition","volume":"478","author":"Goyal","year":"2022","journal-title":"Proc. R. Soc. A"},{"key":"B39","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2103.01197","article-title":"Coordination among neural modules through a shared global workspace","author":"Goyal","year":"2021","journal-title":"arXiv"},{"key":"B40","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1909.10893","article-title":"Recurrent independent mechanisms","author":"Goyal","year":"2019","journal-title":"arXiv"},{"key":"B41","doi-asserted-by":"publisher","first-page":"60","DOI":"10.3389\/frobt.2017.00060","article-title":"The attention schema theory: a foundation for engineering artificial consciousness","volume":"4","author":"Graziano","year":"2017","journal-title":"Front. Robot. AI"},{"key":"B42","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2010.02193","article-title":"Mastering atari with discrete world models","author":"Hafner","year":"2020","journal-title":"arXiv"},{"key":"B43","doi-asserted-by":"publisher","first-page":"1735","DOI":"10.1162\/neco.1997.9.8.1735","article-title":"Long short-term memory","volume":"9","author":"Hochreiter","year":"1997","journal-title":"Neural Comput."},{"key":"B44","first-page":"1","article-title":"Cleanrl: high-quality single-file implementations of deep reinforcement learning algorithms","volume":"23","author":"Huang","year":"2022","journal-title":"J. Mach. Learn. Res"},{"key":"B45","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2107.14795","article-title":"Perceiver IO: a general architecture for structured inputs & outputs","author":"Jaegle","year":"","journal-title":"arXiv"},{"key":"B46","first-page":"4651","article-title":"\u201cPerceiver: general perception with iterative attention,\u201d","volume-title":"Proceedings of the 38th International Conference on Machine Learning","author":"Jaegle","year":""},{"key":"B47","article-title":"On the link between conscious function and general intelligence in humans and machines","author":"Juliani","year":"","journal-title":"Trans. Mach. Learn. Res"},{"key":"B48","article-title":"\u201cThe perceiver architecture is a functional global workspace,\u201d","volume-title":"Proceedings of the Annual Meeting of the Cognitive Science Society","author":"Juliani","year":""},{"key":"B49","doi-asserted-by":"publisher","first-page":"niz016","DOI":"10.1093\/nc\/niz016","article-title":"Information generation as a functional basis of consciousness","volume":"2019","author":"Kanai","year":"2019","journal-title":"Neurosci. Conscious"},{"key":"B50","first-page":"2469","article-title":"\u201cPolicy optimization with demonstrations,\u201d","volume-title":"Proceedings of the 35th International Conference on Machine Learning","author":"Kang","year":"2018"},{"key":"B51","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1412.6980","article-title":"Adam: a method for stochastic optimization","author":"Kingma","year":"2014","journal-title":"arXiv"},{"key":"B52","article-title":"Imagenet classification with deep convolutional neural networks","author":"Krizhevsky","year":"2012","journal-title":"Adv. Neural Inf. Process. Syst"},{"key":"B53","doi-asserted-by":"publisher","first-page":"494","DOI":"10.1016\/j.tics.2006.09.001","article-title":"Towards a true neural stance on consciousness","volume":"10","author":"Lamme","year":"2006","journal-title":"Trends Cogn. Sci"},{"key":"B54","doi-asserted-by":"publisher","first-page":"204","DOI":"10.1080\/17588921003731586","article-title":"How neuroscience will change our view on consciousness","volume":"1","author":"Lamme","year":"2010","journal-title":"Cogn. Neurosci"},{"key":"B55","doi-asserted-by":"publisher","first-page":"365","DOI":"10.1016\/j.tics.2011.05.009","article-title":"Empirical support for higher-order theories of conscious awareness","volume":"15","author":"Lau","year":"2011","journal-title":"Trends Cogn. Sci"},{"key":"B56","first-page":"3361","article-title":"\u201cConvolutional networks for images, speech, and time series,\u201d","author":"LeCun","year":"1995","journal-title":"The Handbook of Brain Theory and Neural Networks"},{"key":"B57","doi-asserted-by":"publisher","first-page":"116059","DOI":"10.1016\/j.neuroimage.2019.116059","article-title":"Interpretable, highly accurate brain decoding of subtly distinct brain states from functional MRI using intrinsic functional networks and long short-term memory recurrent neural networks","volume":"202","author":"Li","year":"2019","journal-title":"Neuroimage"},{"key":"B58","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2112.02027","article-title":"Divergent representations of ethological visual inputs emerge from supervised, unsupervised, and reinforcement learning","author":"Lindsay","year":"2021","journal-title":"arXiv"},{"key":"B59","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1312.5602","article-title":"Playing atari with deep reinforcement learning","author":"Mnih","year":"2013","journal-title":"arXiv"},{"key":"B60","doi-asserted-by":"publisher","DOI":"10.1101\/585760","article-title":"The shift to life on land selected for planning","author":"Mugan","year":"2019","journal-title":"bioRxiv"},{"key":"B61","year":"2018","journal-title":"OpenAI Five"},{"key":"B62","doi-asserted-by":"publisher","first-page":"20130208","DOI":"10.1098\/rstb.2013.0208","article-title":"The neural subjective frame: from bodily signals to perceptual consciousness","volume":"369","author":"Park","year":"2014","journal-title":"Philos. Trans. R. Soc. B: Biol. Sci"},{"key":"B63","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv:2210.13383","article-title":"Evaluating long-term memory in 3D mazes","author":"Pasukonis","year":"2022","journal-title":"arXiv"},{"key":"B64","doi-asserted-by":"publisher","first-page":"109","DOI":"10.1016\/j.neucom.2007.08.001","article-title":"Monophonic sound source separation with an unsupervised network of spiking neurones","volume":"71","author":"Pichevar","year":"2007","journal-title":"Neurocomputing"},{"key":"B65","first-page":"13908","article-title":"Explaining V1 properties with a biologically constrained deep learning architecture","volume":"36","author":"Pogoncheff","year":"2023","journal-title":"Adv. Neural Inf. Process. Syst"},{"key":"B66","article-title":"Alvinn: an autonomous land vehicle in a neural network","author":"Pomerleau","year":"1988","journal-title":"Adv. Neural Inf. Process. Syst"},{"key":"B67","first-page":"1","article-title":"Stable baselines 3: reliable reinforcement learning implementations","volume":"22","author":"Raffin","year":"2021","journal-title":"J. Mach. Learn. Res"},{"key":"B68","volume-title":"Principles of Neurodynamics: Perceptrons and the Theory of Brain Mechanisms, Vol. 55","author":"Rosenblatt","year":"1962"},{"key":"B69","doi-asserted-by":"publisher","first-page":"155","DOI":"10.1080\/09515089308573085","article-title":"Higher-order thoughts and the appendage theory of consciousness","volume":"6","author":"Rosenthal","year":"1993","journal-title":"Philos. Psychol"},{"key":"B70","doi-asserted-by":"publisher","first-page":"26","DOI":"10.7551\/mitpress\/5236.001.0001","article-title":"\u201cA general framework for parallel distributed processing,\u201d","author":"Rumelhart","year":"1986","journal-title":"Parallel Distributed Processing: Explorations in the Microstructure of Cognition, vol. 1"},{"key":"B71","doi-asserted-by":"publisher","first-page":"211","DOI":"10.1007\/s11263-015-0816-y","article-title":"Imagenet large scale visual recognition challenge","volume":"115","author":"Russakovsky","year":"2015","journal-title":"Int. J. Comput. Vis"},{"key":"B72","doi-asserted-by":"publisher","first-page":"338","DOI":"10.21437\/Interspeech.2014-80","article-title":"Long short-term memory recurrent neural network architectures for large scale acoustic modeling","volume":"2014","author":"Sak","year":"2014","journal-title":"Proc. Interspeech"},{"key":"B73","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1707.06347","article-title":"Proximal policy optimization algorithms","author":"Schulman","year":"2017","journal-title":"arXiv"},{"key":"B74","doi-asserted-by":"publisher","first-page":"1344","DOI":"10.1167\/17.10.1344","article-title":"Comparing human and convolutional neural network performance on scene segmentation","volume":"17","author":"Seijdel","year":"2017","journal-title":"J. Vis"},{"key":"B75","doi-asserted-by":"publisher","first-page":"433","DOI":"10.1016\/j.concog.2005.11.005","article-title":"A cognitive architecture that combines internal simulation with a global workspace","volume":"15","author":"Shanahan","year":"2006","journal-title":"Conscious. Cogn"},{"key":"B76","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv:1409.1556","article-title":"Very deep convolutional networks for large-scale image recognition","author":"Simonyan","year":"2014","journal-title":"arXiv"},{"key":"B77","article-title":"\u201cThe neural MMO platform for massively multiagent research,\u201d","author":"Suarez","year":"2021","journal-title":"Proceedings of the Neural Information Processing Systems Track on Datasets and Benchmarks, Vol. 1"},{"key":"B78","volume-title":"Reinforcement Learning: An Introduction","author":"Sutton","year":"2018"},{"key":"B79","doi-asserted-by":"publisher","first-page":"222010","DOI":"10.3389\/fnins.2016.00524","article-title":"Computational models of auditory scene analysis: a review","volume":"10","author":"Szab\u00f3","year":"2016","journal-title":"Front. Neurosci"},{"key":"B80","doi-asserted-by":"publisher","first-page":"692","DOI":"10.1016\/j.tins.2021.04.005","article-title":"Deep learning and the global workspace theory","volume":"44","author":"VanRullen","year":"2021","journal-title":"Trends Neurosci"},{"key":"B81","article-title":"\u201cAttention is all you need,\u201d","author":"Vaswani","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"key":"B82","doi-asserted-by":"publisher","first-page":"e2102421118","DOI":"10.1073\/pnas.2102421118","article-title":"The attention schema theory in a neural network agent: controlling visuospatial attention using a descriptive model of attention","volume":"118","author":"Wilterson","year":"2021","journal-title":"Proc. Nat. Acad. Sci"},{"key":"B83","doi-asserted-by":"publisher","first-page":"101844","DOI":"10.1016\/j.pneurobio.2020.101844","article-title":"Attention control and the attention schema theory of consciousness","volume":"195","author":"Wilterson","year":"2020","journal-title":"Prog. Neurobiol"},{"key":"B84","unstructured":"YoonJ.\n          dreamer-torch2023"},{"key":"B85","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2208.12345","article-title":"Light-weight probing of unsupervised representations for reinforcement learning","author":"Zhang","year":"2022","journal-title":"arXiv"}],"container-title":["Frontiers in Computational Neuroscience"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fncom.2024.1352685\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,6,14]],"date-time":"2024-06-14T05:18:05Z","timestamp":1718342285000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fncom.2024.1352685\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,6,14]]},"references-count":85,"alternative-id":["10.3389\/fncom.2024.1352685"],"URL":"https:\/\/doi.org\/10.3389\/fncom.2024.1352685","relation":{},"ISSN":["1662-5188"],"issn-type":[{"value":"1662-5188","type":"electronic"}],"subject":[],"published":{"date-parts":[[2024,6,14]]},"article-number":"1352685"}}