The Next Decade in AI: Four Steps Towards Robust Artificial Intelligence

Abstract

Recent research in artificial intelligence and machine learning has largely emphasized general-purpose learning and ever-larger training sets and more and more compute. In contrast, I propose a hybrid, knowledge-driven, reasoning-based approach, centered around cognitive models, that could provide the substrate for a richer, more robust AI than is currently possible.

5 figures

References in corpus (26)

• Solving Rubik's Cube with a Robot Hand
• Large-scale Simple Question Answering with Memory Networks
• Deep Learning: A Critical Appraisal
• Adversarial Examples for Evaluating Reading Comprehension Systems
• Towards a Human-like Open-Domain Chatbot
• MONet: Unsupervised Scene Decomposition and Representation
• Differentiable Learning of Logical Rules for Knowledge Base Reasoning
• Schema Networks: Zero-shot Transfer with a Generative Causal Model of Intuitive Physics
• Tracking the World State with Recurrent Entity Networks
• Unsupervised Predictive Memory in a Goal-Directed Agent
• Logic Tensor Networks: Deep Learning and Logical Reasoning from Data and Knowledge
• A Meta-Transfer Objective for Learning to Disentangle Causal Mechanisms
• Compositional Semantic Parsing on Semi-Structured Tables
• Frequently Asked Questions for: The Atoms of Neural Computation
• Recurrent Independent Mechanisms
• Adversarial NLI: A New Benchmark for Natural Language Understanding
• Machine Theory of Mind
• Environmental drivers of systematicity and generalization in a situated agent
• Deep Learning for Symbolic Mathematics
• Reasoning About Physical Interactions with Object-Oriented Prediction and Planning
• Enhancing the Transformer with Explicit Relational Encoding for Math Problem Solving
• Entity Abstraction in Visual Model-Based Reinforcement Learning
• Learning Symbolic Physics with Graph Networks
• From 'F' to 'A' on the N.Y. Regents Science Exams: An Overview of the Aristo Project
• The Use of Deep Learning for Symbolic Integration: A Review of (Lample and Charton, 2019)
• Learning dynamic polynomial proofs

Cited by in corpus (44)

• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
• Out of One, Many: Using Language Models to Simulate Human Samples
• Deep Learning Based Text Classification: A Comprehensive Review
• Discovering Symbolic Models from Deep Learning with Inductive Biases
• Designing Interpretable ML System to Enhance Trust in Healthcare: A Systematic Review to Proposed Responsible Clinician-AI-Collaboration Framework
• A Survey on Hyperdimensional Computing aka Vector Symbolic Architectures, Part II: Applications, Cognitive Models, and Challenges
• Trustworthy AI and Robotics and the Implications for the AEC Industry: A Systematic Literature Review and Future Potentials
• Graph Neural Networks Meet Neural-Symbolic Computing: A Survey and Perspective
• Neural Collaborative Reasoning
• The effect of source disclosure on evaluation of AI-generated messages: A two-part study
• Enhancing Knowledge Retrieval with In-Context Learning and Semantic Search through Generative AI
• Known Operator Learning and Hybrid Machine Learning in Medical Imaging -- A Review of the Past, the Present, and the Future
• Discrete and continuous representations and processing in deep learning: Looking forward
• A Review of Emerging Research Directions in Abstract Visual Reasoning
• A Review of Winograd Schema Challenge Datasets and Approaches
• The Switch, the Ladder, and the Matrix: Models for Classifying AI Systems
• PLATO-XL: Exploring the Large-scale Pre-training of Dialogue Generation
• Robust Conversational AI with Grounded Text Generation
• Online Deep Learning from Doubly-Streaming Data
• Benchmark for Models Predicting Human Behavior in Gap Acceptance Scenarios
• Building Human-like Communicative Intelligence: A Grounded Perspective
• Commonsense Knowledge in Wikidata
• "Weak AI" is Likely to Never Become "Strong AI", So What is its Greatest Value for us?
• Consolidating Commonsense Knowledge
• Trusted Artificial Intelligence: Towards Certification of Machine Learning Applications
• Unsupervised Dialogue Topic Segmentation in Hyperdimensional Space
• Symbolic Behaviour in Artificial Intelligence
• A World-Self Model Towards Understanding Intelligence
• Conversational Answer Generation and Factuality for Reading Comprehension Question-Answering
• Stochastic Adversarial Gradient Embedding for Active Domain Adaptation
• Design considerations for a hierarchical semantic compositional framework for medical natural language understanding
• Extending Automated Deduction for Commonsense Reasoning
• Generalisation in Neural Networks Does not Require Feature Overlap
• Target Consistency for Domain Adaptation: when Robustness meets Transferability
• Decrypting Cryptic Crosswords: Semantically Complex Wordplay Puzzles as a Target for NLP
• Learning Algebraic Representation for Systematic Generalization in Abstract Reasoning
• The curious case of developmental BERTology: On sparsity, transfer learning, generalization and the brain
• Towards an Interface Description Template for AI-enabled Systems
• Attention-based Contrastive Learning for Winograd Schemas
• Knowledge as Invariance -- History and Perspectives of Knowledge-augmented Machine Learning
• The case for psychometric artificial general intelligence
• Understanding in Artificial Intelligence
• AI Centered on Scene Fitting and Dynamic Cognitive Network
• Towards Zero-shot Commonsense Reasoning with Self-supervised Refinement of Language Models