Open-Endedness is Essential for Artificial Superhuman Intelligence

TL;DR

This paper argues that open-endedness is crucial for developing artificial superhuman intelligence, proposing formal definitions and pathways using foundation models to achieve continuous novelty and learnability, with significant safety considerations.

Contribution

It introduces a formal definition of open-endedness, advocates its role in ASI development, and outlines a pathway using foundation models to foster continuous novelty and discovery.

Findings

Open-endedness is essential for ASI development.

Foundation models can be built to support open-ended, self-improving AI.

Open-ended AI has important safety implications.

Abstract

In recent years there has been a tremendous surge in the general capabilities of AI systems, mainly fuelled by training foundation models on internetscale data. Nevertheless, the creation of openended, ever self-improving AI remains elusive. In this position paper, we argue that the ingredients are now in place to achieve openendedness in AI systems with respect to a human observer. Furthermore, we claim that such open-endedness is an essential property of any artificial superhuman intelligence (ASI). We begin by providing a concrete formal definition of open-endedness through the lens of novelty and learnability. We then illustrate a path towards ASI via open-ended systems built on top of foundation models, capable of making novel, humanrelevant discoveries. We conclude by examining the safety implications of generally-capable openended AI. We expect that open-ended foundation models will prove to be an increasingly fertile and safety-critical area of research in the near future.