About this Research Topic
Artificial Intelligence (AI) is the study of techniques that allow machines to learn, reason, and act to achieve goals. Much of the current AI research focuses on creating narrow AI systems that perform well-defined tasks such as facial recognition, internet search, or driving a car. The long-term goal of AI research is to develop general AI with human-like cognitive capabilities. Whereas narrow AI performs a single task, general AI would act similarly to humans at nearly every cognitive task of interest.
Machine learning is an essential type of narrow AI. It permits machines to learn from big data sets without being explicitly programmed. Current AI research focuses on a kind of machine learning called deep learning, a family of statistical techniques for classifying patterns using artificial neural networks (ANNs) with many hidden layers. We have seen breakthroughs in image and speech recognition, language translation, navigation, and game playing. However, even though deep learning outperforms humans in isolated domains, AI researchers increasingly worry that the current path will not lead to machines with human-like intelligence.
Can neuroscience contribute to AI research? The human brain was the initial inspiration for ANNs, but we know that biology is significantly more complex than current artificial systems. Biological neurons have complex dendritic structures and are connected in complex ways. Most AI systems based on ANNs require separate, offline training whereas much of the learning in the brain occurs continuously using streaming sensory data. Plasticity in the brain is also more nuanced, occurring at many different time scales and affected by a variety of neuromodulators. Finally, most learning methods in AI ignore sensorimotor integration.
It is possible that AI researchers could significantly benefit from neuroscience and develop AI models that are directly based on the computational principles of the brain. This is not an easy task. Neuroscience papers deal with diverse aspects of the brain, have their own terminology, and often deal with issues that are irrelevant to AI. It is difficult for AI researchers to parse through these papers and select relevant results.
The goal of this Research Topic is to make the core ideas accessible to both AI researchers and neuroscientists. We solicit original research and review papers that cater to a broad readership. Given the interdisciplinary nature of this topic, the ideal article will be accessible to any researcher interested in biologically-constrained general AI.
A successful paper could address one or more of the areas and questions:
• What are the limitations of narrow AI, and how can neuroscience help remove those limitations?
• Novel neuroscience-based approaches to continuous and lifelong learning.
• How can AI benefit from the latest discoveries in sensory systems, including vision, audition, and somatosensation?
• The neuroscience behind sensorimotor integration, navigation, and motor control, and their relationship to AI.
• The neuroscience behind cognitive maps, decision making, and abstract reasoning, and how they might impact AI.
• Can the neuroscience of emotions help create general AI?
• How can neuroscience help AI researchers create “common sense”?
• What can AI researchers learn from developmental neuroscience and the development of young children?
Topic Editor Subutai Ahmad is employed by the company Numenta and Topic Editor Adam Henry Marblestone is employed by the company DeepMind. All other Topic Editors declare no competing interests with regards to the Research Topic subject.
Machine learning is an essential type of narrow AI. It permits machines to learn from big data sets without being explicitly programmed. Current AI research focuses on a kind of machine learning called deep learning, a family of statistical techniques for classifying patterns using artificial neural networks (ANNs) with many hidden layers. We have seen breakthroughs in image and speech recognition, language translation, navigation, and game playing. However, even though deep learning outperforms humans in isolated domains, AI researchers increasingly worry that the current path will not lead to machines with human-like intelligence.
Can neuroscience contribute to AI research? The human brain was the initial inspiration for ANNs, but we know that biology is significantly more complex than current artificial systems. Biological neurons have complex dendritic structures and are connected in complex ways. Most AI systems based on ANNs require separate, offline training whereas much of the learning in the brain occurs continuously using streaming sensory data. Plasticity in the brain is also more nuanced, occurring at many different time scales and affected by a variety of neuromodulators. Finally, most learning methods in AI ignore sensorimotor integration.
It is possible that AI researchers could significantly benefit from neuroscience and develop AI models that are directly based on the computational principles of the brain. This is not an easy task. Neuroscience papers deal with diverse aspects of the brain, have their own terminology, and often deal with issues that are irrelevant to AI. It is difficult for AI researchers to parse through these papers and select relevant results.
The goal of this Research Topic is to make the core ideas accessible to both AI researchers and neuroscientists. We solicit original research and review papers that cater to a broad readership. Given the interdisciplinary nature of this topic, the ideal article will be accessible to any researcher interested in biologically-constrained general AI.
A successful paper could address one or more of the areas and questions:
• What are the limitations of narrow AI, and how can neuroscience help remove those limitations?
• Novel neuroscience-based approaches to continuous and lifelong learning.
• How can AI benefit from the latest discoveries in sensory systems, including vision, audition, and somatosensation?
• The neuroscience behind sensorimotor integration, navigation, and motor control, and their relationship to AI.
• The neuroscience behind cognitive maps, decision making, and abstract reasoning, and how they might impact AI.
• Can the neuroscience of emotions help create general AI?
• How can neuroscience help AI researchers create “common sense”?
• What can AI researchers learn from developmental neuroscience and the development of young children?
Topic Editor Subutai Ahmad is employed by the company Numenta and Topic Editor Adam Henry Marblestone is employed by the company DeepMind. All other Topic Editors declare no competing interests with regards to the Research Topic subject.
Keywords: sensorimotor integration, embodied reasoning, neuro-inspired AI, meta-learning, plasticity mechanisms
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
