Work

Replicating cognitive self-sufficiency to realize computational self-sufficiency in Artificial General Intelligence models

Year: 2023
Type: dissertation
Abstract: There is compelling evidence that qualitative states are integral to realizing the type of autonomy exhibited in cognitive general intelligence.Cognitive agents leverage experiential qualitative states in forming knowledge of the world, adopting attitudes to the world, and processing responses to general problems in the world.This capacity for self-directed, adaptive problem-solving, that allows cognitive agents to solve both familiar and novel problems, serves as the main inspiration for Artificial General Intelligence (AGI).However, current AGI development focuses on computational models that largely ignore the role of qualitative states.If the goal of AGI is to realize artefacts with general intelligence that is functionally like cognitive general intelligence, then it should be critical to replicate the selfdirected autonomy that qualitative states enable in cognitive general intelligence.This research aims to resolve the question 'How can the roles of qualitative states in supporting the self-directedness of cognitive general intelligence be replicated in models of artificial general intelligence?'The thesis identifies that qualitative states are pivotal in (1) facilitating the self-grounding of mental computation with underived meaning from the world.(2) Enabling self-creditable control of the computation that yields actions.These two roles are critical for accounting for the intentional actions that are the hallmark of the autonomy exercised by cognitive intelligence.The thesis delineates natural accounts of these two functional roles. From these roles, it extracts two replicable principles: the Analog Principle and the InternalModel Principle.These principles can be used to ameliorate computational models of intelligence with autonomy like cognitive general intelligence.To demonstrate this, the thesis describes a model of intelligence supported by these principles and assesses the model for the intentional agency that is definitive of cognitive general intelligence.
Institution: University of York
Language: English
Open Access status: green
OpenAlex ID: W7024352418
Cites: 147
Cited by:
Related to: 10