Deep Learning and Artificial General Intelligence: Still a Long Way to Go

TL;DR

This paper critically examines the potential of deep learning to achieve Artificial General Intelligence, highlighting five major reasons why current deep neural networks are not yet suitable for this goal.

Contribution

It provides a critical analysis of deep neural networks, identifying key limitations preventing their use as the foundation for AGI.

Findings

Deep neural networks lack generalization capabilities.

Current architectures do not support reasoning or understanding.

Significant advancements are needed before deep learning can enable AGI.

Abstract

In recent years, deep learning using neural network architecture, i.e. deep neural networks, has been on the frontier of computer science research. It has even lead to superhuman performance in some problems, e.g., in computer vision, games and biology, and as a result the term deep learning revolution was coined. The undisputed success and rapid growth of deep learning suggests that, in future, it might become an enabler for Artificial General Intelligence (AGI). In this article, we approach this statement critically showing five major reasons of why deep neural networks, as of the current state, are not ready to be the technique of choice for reaching AGI.