Levels of AGI: Operationalising Progress on the Path to AGI
Meredith Ringel Morris et al. 2023. (View Paper → )
We propose a framework for classifying the capabilities and behavior of Artificial General Intelligence(AGI) models and their precursors. This framework introduces levels of AGI performance, generality, and autonomy. It is our hope that this framework will be useful in an analogous way to the levels of autonomous driving, by providing a common language to compare models, assess risks, and measure progress along the path to AGI.
The authors present a structured approach to understanding and categorizing progress towards Artificial General Intelligence (AGI), which has proven to be a challenging area due to its broad and often ambiguous definitions.
Defining AGI is a real challenge and is a subject of much debate in the field. This paper begins by summarizing nine popular views on how to define AGI, including the Turing Test, and discusses their pros and cons. It then extracts six principles from these definitions to move forward.
The paper introduces five levels of AGI (No AI, Emerging, Competent, Expert, Virtuoso, Superhuman) based on percentile performance compared to skilled adults in a set of tasks that are not defined. It provides a clear distinction between Narrow and General AI at each level, with examples.
Additionally, the paper introduces levels of autonomy, which offers an interesting perspective that I had not considered before. These levels include No AI, AI as a tool, AI as a consultant, AI as a collaborator, AI as an expert, and AI as an agent.
If you are interested in AGI, this paper is highly recommended.