What is consciousness, and could machines have it?

Area of Science:

• Neuroscience
• Artificial Intelligence
• Cognitive Science
• Philosophy of Mind

Background:

• The debate on machine consciousness hinges on understanding biological consciousness in the human brain.
• Current artificial intelligence successes do not equate to genuine consciousness.
• The term "consciousness" is often used ambiguously, encompassing distinct cognitive functions.

Purpose of the Study:

• To differentiate between distinct types of information processing in the human brain related to consciousness.
• To analyze current machine learning architectures in the context of human unconscious and conscious processing.
• To propose a framework for developing more sophisticated AI inspired by neural mechanisms of consciousness.

Main Methods:

• Review of psychological and neural science literature on unconscious (C0) and conscious (C1, C2) brain computations.
• Conceptual analysis of information processing in human consciousness and current AI systems.
• Comparative analysis of computational mechanisms underlying C0, C1, and C2 processes.

Main Results:

• Consciousness is proposed to involve two distinct computations: global information broadcasting (C1) and self-monitoring of computations (C2).
• Current AI predominantly implements computations analogous to unconscious processing (C0) in the human brain.
• A clear distinction is drawn between unconscious processing (C0), global availability (C1), and subjective certainty (C2).

Conclusions:

• Understanding the neural basis of C1 and C2 is crucial for advancing AI towards genuine consciousness.
• Future AI architectures could be inspired by the distinct computational principles of human C0, C1, and C2 processes.
• This research provides a neuroscientific framework for evaluating and developing artificial consciousness.