Reconciling deep learning with symbolic artificial intelligence: representing objects and relations

2019/01/04 by Marta Garnelo, Murray Shanahan · 1 citation
Computer Science · Social Sciences · Psychology · Mathematics · #Topic Modeling #Language and cultural evolution #Neural Networks and Applications #Connectionism #Artificial intelligence #Computer science #Field (mathematics) #Symbolic artificial intelligence #Deep learning #Key (lock) #Strengths and weaknesses #Cognitive science #Artificial neural network #The Symbolic #Artificial Intelligence System #Epistemology #Psychology #Mathematics #Philosophy

paper · pdf · doi:10.1016/j.cobeha.2018.12.010

openalex publication_date 2019/01/04 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/01

Abstract

In the history of the quest for human-level artificial intelligence, a number of rival paradigms have vied for supremacy. Symbolic artificial intelligence was dominant for much of the 20th century, but currently a connectionist paradigm is in the ascendant, namely machine learning with deep neural networks. However, both paradigms have strengths and weaknesses, and a significant challenge for the field today is to effect a reconciliation. A central tenet of the symbolic paradigm is that intelligence results from the manipulation of abstract compositional representations whose elements stand for objects and relations. If this is correct, then a key objective for deep learning is to develop architectures capable of discovering objects and relations in raw data, and learning how to represent them in ways that are useful for downstream processing. This short review highlights recent progress in this direction.

Citations

• Show, Attend and Tell: Neural Image Caption Generation with Visual Attention
• Stochastic Backpropagation and Approximate Inference in Deep Generative Models
• Auto-Encoding Variational Bayes
• Deep Learning in Neural Networks: An Overview
• Human-level control through deep reinforcement learning
• End-to-End Training of Deep Visuomotor Policies
• Representation Learning: A Review and New Perspectives
• Mastering the game of Go with deep neural networks and tree search
• Neuro-Symbolic Program Synthesis
• Interaction Networks for Learning about Objects, Relations and Physics
• Inducing Interpretable Representations with Variational Autoencoders
• A Compositional Object-Based Approach to Learning Physical Dynamics
• Deep Learning: A Critical Appraisal
• Disentangling by Factorising
• SCAN: Learning Hierarchical Compositional Visual Concepts
• Measuring abstract reasoning in neural networks
• Human-like generalization in a machine through predicate learning
• Relational Deep Reinforcement Learning
• InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets
• Exploiting Generative AI to Scale up Intelligent Tutoring Systems

Cited by

• Symbolic artificial intelligence [wikipedia]

Related

• Learning like humans with Deep Symbolic Networks
• Emergence of Machine Language: Towards Symbolic Intelligence with Neural Networks
• Neural-Symbolic Computing: An Effective Methodology for Principled Integration of Machine Learning and Reasoning
• Symbolic Techniques for Deep Learning: Challenges and Opportunities
• Neural-Symbolic Learning and Reasoning: Contributions and Challenges