Artificial intelligence, human cognition, and conscious supremacy

Area of Science:

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

Background:

• Consciousness, defined by phenomenal properties like qualia and self-awareness, is contrasted with unconscious states.
• The success of large language models (LLMs) prompts re-evaluation of the relationship between intelligence and consciousness.
• Biological systems often show intelligence and consciousness together, but exceptions suggest possible dissociations.

Purpose of the Study:

• To explore the computational significance of human conscious processing.
• To identify cognitive domains potentially unique to consciousness.
• To discuss the implications for artificial intelligence alignment.

Main Methods:

• Review of existing literature on consciousness and computation.
• Analysis of cognitive tasks potentially unique to conscious processing.
• Conceptual proposal of 'conscious supremacy' analogous to quantum supremacy.

Main Results:

• Several cognitive domains, including flexible attention, novel context handling, and integrated sensory processing, may be unique to consciousness.
• AI systems like LLMs demonstrate intelligence potentially without consciousness, unlike typical human processing.
• A distinction is proposed between computations requiring consciousness and those that can be unconscious.

Conclusions:

• Consciousness may play a distinct computational role, particularly in flexible and adaptive cognition.
• The concept of 'conscious supremacy' could guide the search for consciousness-specific computations.
• Understanding these distinctions is crucial for aligning AI and human cognitive processes.