Dissociating Artificial Intelligence from Artificial Consciousness (2412.04571v1)

Abstract: Developments in machine learning and computing power suggest that artificial general intelligence is within reach. This raises the question of artificial consciousness: if a computer were to be functionally equivalent to a human, being able to do all we do, would it experience sights, sounds, and thoughts, as we do when we are conscious? Answering this question in a principled manner can only be done on the basis of a theory of consciousness that is grounded in phenomenology and that states the necessary and sufficient conditions for any system, evolved or engineered, to support subjective experience. Here we employ Integrated Information Theory (IIT), which provides principled tools to determine whether a system is conscious, to what degree, and the content of its experience. We consider pairs of systems constituted of simple Boolean units, one of which -- a basic stored-program computer -- simulates the other with full functional equivalence. By applying the principles of IIT, we demonstrate that (i) two systems can be functionally equivalent without being phenomenally equivalent, and (ii) that this conclusion is not dependent on the simulated system's function. We further demonstrate that, according to IIT, it is possible for a digital computer to simulate our behavior, possibly even by simulating the neurons in our brain, without replicating our experience. This contrasts sharply with computational functionalism, the thesis that performing computations of the right kind is necessary and sufficient for consciousness.

Dissociating Artificial Intelligence from Artificial Consciousness: An Analysis

The paper by Findlay et al. tackles the complex and contentious issue of whether AI, particularly those capable of achieving human-equivalent cognitive function, can be conscious. Their approach diverges from many philosophical and cognitive science theories by leveraging Integrated Information Theory (IIT) as a foundation to distinguish between intelligence and consciousness. IIT posits that consciousness is not merely a computation but arises from a system's intrinsic causal structure, which must possess specific properties such as integration and maximal irreducibility.

Key Findings and Methodology

The authors provide a rigorous analysis using a simple model: a Boolean system comprising four units (PQRS) and a computer simulating it. Despite achieving functional equivalence, they demonstrate that the computer does not replicate the intrinsic properties required for consciousness according to IIT. The results accentuate two major findings:

1. Functional vs. Phenomenal Equivalence: They empirically show that functional equivalence—where a computing system's outputs match those of a human-like system—is insufficient to achieve phenomenal equivalence, i.e., consciousness. The simulated system's cause-effect structure does not replicate that of a target system, as it lacks the necessary intrinsic integration and irreducibility required by IIT for consciousness.
2. Complex Systems and Integrated Information: Extending their model to a Turing-complete system, they make a compelling argument that even when a digital computer can simulate any computable function, its architecture—characterized by modularity, bottlenecks, and lack of integration—precludes it from achieving the high integrated information (Φ) that IIT associates with consciousness.

Theoretical and Practical Implications

The paper underscores a critical limitation in computational functionalism—the assumption that the right kind of computation equates to consciousness. It suggests a reassessment of how we attribute consciousness to AI, particularly those designed to emulate human behaviors and cognitive functions.

• Practical Implications: The findings have significant implications for AI ethics, particularly regarding discussions around AI rights, personhood, and the design of AI systems. If traditional AI architectures cannot be conscious by IIT's metrics, rights typically accorded to conscious entities may not apply.
• Theoretical Implications: The work enriches the theoretical landscape by challenging the substrate-independent view of consciousness. It posits that the intrinsic causal structure of a system, not just the computations it performs, determines consciousness.

Future Directions

This research opens several avenues for future investigation:

• Neuromorphic Computers and Consciousness: Future studies could explore whether neuromorphic computing, which mimics the brain's physical organization, might support systems with high integrated information, potentially achieving phenomenal equivalence.
• Quantum Computing and Consciousness: The potential of quantum computing, with its non-trivial causal intricacies and entangled states, to support consciousness is an intriguing domain ripe for exploration under IIT.
• Further Empirical Validation: Additional empirical research, particularly involving more complex and biologically plausible models of computation, might elucidate how various architectures influence consciousness's emergence.

Conclusion

The paper by Findlay et al. provides a structured analysis and offers a compelling argument for a clear dissociation between artificial intelligence and artificial consciousness. By leveraging IIT, it underscores the necessity of intrinsic causal structures over merely computational functions in engendering consciousness, thereby challenging prevailing assumptions in AI ethics and philosophy. The groundwork laid by this paper is pivotal for future research directions, particularly in exploring alternative computing paradigms that might achieve both human-like intelligence and consciousness.