The Next Decade in AI: Four Steps Towards Robust Artificial Intelligence (2020-02-14T00:00:00.000000Z)

TL;DR

This work proposes 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.

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.

Authors

G. Marcus

2 papers

TL;DR

Abstract

Authors

References111 items

MEMO: A Deep Network for Flexible Combination of Episodic Memories

Towards a Human-like Open-Domain Chatbot

Concepts and Compositionality: In Search of the Brain's Language of Thought.

Dendritic action potentials and computation in human layer 2/3 cortical neurons

Differentiable Reasoning on Large Knowledge Bases and Natural Language

The Use of Deep Learning for Symbolic Integration: A Review of (Lample and Charton, 2019)

Neural Module Networks for Reasoning over Text

Deep Learning for Symbolic Mathematics

Mastering Atari, Go, chess and shogi by planning with a learned model

Linguistic Fundamentals for Natural Language Processing II: 100 Essentials from Semantics and Pragmatics

Memory Augmented Recursive Neural Networks

Adversarial NLI: A New Benchmark for Natural Language Understanding

Entity Abstraction in Visual Model-Based Reinforcement Learning

Solving Rubik's Cube with a Robot Hand

Emergent Systematic Generalization in a Situated Agent

Environmental drivers of systematicity and generalization in a situated agent

Enhancing the Transformer with Explicit Relational Encoding for Math Problem Solving

Recurrent Independent Mechanisms

Learning Symbolic Physics with Graph Networks

From 'F' to 'A' on the N.Y. Regents Science Exams: An Overview of the Aristo Project

The Book of Why: The New Science of Cause and Effect†

GENESIS: Generative Scene Inference and Sampling with Object-Centric Latent Representations

Learning higher-order logic programs

Shaping Belief States with Generative Environment Models for RL

Learning dynamic polynomial proofs

The Neuro-Symbolic Concept Learner: Interpreting Scenes, Words, and Sentences From Natural Supervision

Integrating New Knowledge into a Neural Network without Catastrophic Interference: Computational and Theoretical Investigations in a Hierarchically Structured Environment

A Meta-Transfer Objective for Learning to Disentangle Causal Mechanisms

MONet: Unsupervised Scene Decomposition and Representation

Strike (With) a Pose: Neural Networks Are Easily Fooled by Strange Poses of Familiar Objects

Pyro: Deep Universal Probabilistic Programming

Reasoning About Physical Interactions with Object-Oriented Prediction and Planning

Learning Finite State Representations of Recurrent Policy Networks

The Book of Why: The New Science of Cause and Effect

Unsupervised Predictive Memory in a Goal-Directed Agent

World Models

Machine Theory of Mind

Deep Learning: A Critical Appraisal

A generative vision model that trains with high data efficiency and breaks text-based CAPTCHAs

Precision Learning: Towards Use of Known Operators in Neural Networks

Learning Explanatory Rules from Noisy Data

Still not systematic after all these years: On the compositional skills of sequence-to-sequence recurrent networks

Mastering the game of Go without human knowledge

Steps Toward Robust Artificial Intelligence

Deep Reinforcement Learning that Matters

Adversarial Examples for Evaluating Reading Comprehension Systems

Commonsense reasoning about containers using radically incomplete information

Schema Networks: Zero-shot Transfer with a Generative Causal Model of Intuitive Physics

Attention is All you Need

Differentiable Learning of Logical Rules for Knowledge Base Reasoning

Reasoning in Non-probabilistic Uncertainty: Logic Programming and Neural-Symbolic Computing as Examples

Assembly pointers for variable binding in networks of spiking neurons

Tracking the World State with Recurrent Entity Networks

Comprehensive transcriptional map of primate brain development

A Comparative Evaluation of Approximate Probabilistic Simulation and Deep Neural Networks as Accounts of Human Physical Scene Understanding

Statistical Relational Artificial Intelligence: Logic, Probability, and Computation

Basic Reasoning with Tensor Product Representations

FlashMeta: a framework for inductive program synthesis

Compositional Semantic Parsing on Semi-Structured Tables

Large-scale Simple Question Answering with Memory Networks

Human-level control through deep reinforcement learning

The atoms of neural computation

Transcriptional Landscape of the Prenatal Human Brain

A measure of intelligence

Kluge: The Haphazard Construction of the Human Mind

Memory and the Computational Brain: Why Cognitive Science will Transform Neuroscience

Neural-Symbolic Cognitive Reasoning

Roots, stems, and the universality of lexical representations: Evidence from Hebrew

Markov logic networks

The Organization of Learning

The Birth of the Mind: How a Tiny Number of Genes Creates The Complexities of Human Thought

The scope of linguistic generalizations: evidence from Hebrew word formation

Detecting blickets: how young children use information about novel causal powers in categorization and induction.

The Origin of Concepts

Rethinking Eliminative Connectionism

Hybrid Connectionist-Symbolic Modules: A Report from the IJCAI-95 Workshop on Connectionist-Symbolic Integration