TLDR: A new computational model defines consciousness as an evolutionary advantage for intelligent entities operating on reactive substrates. It proposes a “consciousness switch” that allows distinguishing between real and hypothetical experiences, preventing autonomous physical reactions to imagined threats. The model suggests artificial consciousness is achievable but unnecessary for advanced AI if autonomous responses are absent, challenging traditional views on consciousness and its necessity for AI.
The age-old enigma of consciousness, its precise definition, and the underlying mechanisms that bring it about, has gained renewed urgency with the rapid advancements in artificial intelligence. The scientific community remains divided between two major viewpoints: physicalism, which posits consciousness as a physical process that can be computationally modeled, and natural dualism, which rejects this idea.
A new research paper, “Consciousness, natural and artificial: an evolutionary advantage for reasoning on reactive substrates”, by Warisa Sritriratanarak and Paulo Garcia from Chulalongkorn University, offers a computational model that supports the physicalism hypothesis. The authors argue that a precise model of consciousness, whether natural or artificial, can be achieved by considering the underlying substrate (biological or digital) and accounting for reactive behaviors within its sub-systems, such as autonomous physiological responses.
Consciousness: A Substrate-Dependent Advantage
A key finding of this research is that, unlike most other computational processes, consciousness is not independent of its substrate. Instead, possessing consciousness provides an evolutionary advantage for intelligent entities. Surprisingly, the paper also concludes that while fully artificial consciousness is entirely possible, it is also feasible to create artificial intelligence of any level without consciousness whatsoever, and there is no inherent benefit in imbuing artificial systems with it.
To avoid ambiguity, the researchers define consciousness through two core aspects: awareness of the phenomenal experience – an entity’s sense of “here and now” grounded in its reality – and subjective experience – how two distinct entities with the same cognitive substrate can experience the same stimuli differently (qualia).
The Evolutionary Need for a “Consciousness Switch”
The model is built upon three axioms:
- Ever-evolving perception: An entity’s perception function changes over time, shaped by past experiences.
- Autonomous triggers: The perception function can initiate substrate responses (like fight-or-flight) before higher-level reasoning takes over. These are ancient evolutionary responses.
- Percept of self: An entity’s memory includes a representation of its own physical substrate and its sensing, perception, and reasoning processes.
The paper explains that intelligent entities, when reasoning about hypothetical future states (e.g., “what if I do this?”), engage the same sensory and perceptual sub-systems used for processing real-time input. Without a mechanism to differentiate between real and imagined scenarios, these hypothetical threats could trigger actual autonomous physical responses, leading to “damage caused by thinking.”
This is where consciousness comes in. The authors propose a “consciousness switch” within the substrate machinery. This switch, controlled by the higher-level reasoning process, allows for the dampening or turning off of autonomous responses when processing hypothetical data. This enables an entity to distinguish between reality and imagination, providing a crucial evolutionary advantage by preventing maladaptive physical reactions to non-existent threats. This switch can be interpreted as a boolean value indicating “this is real, happening here and now.”
Subjective Experience and AI Implications
The concept of subjective experience (qualia) is explained by the ever-evolving nature of perception. Since no two entities can occupy the same space at the same time and thus will always have slightly different experiences, their perception functions will necessarily diverge over time, leading to unique subjective interpretations of the same stimuli.
For artificial intelligence, the implications are profound. The “consciousness switch” can be easily implemented in an AI system, making artificial consciousness readily achievable. However, the paper highlights a critical point: if an AI is designed without autonomous, reactive responses tied to its substrate, then there is no functional need for it to be conscious. A conscious AI on a reactive substrate would, at best, perform similarly to a non-conscious AI on a non-reactive substrate.
Also Read:
- The Elusive Quest for Common Sense in Artificial Intelligence: A Deep Dive into Current Limitations
- Navigating Trust: When Should We Defer to AI’s Expertise?
Challenging Philosophical Arguments
The research also addresses long-standing philosophical arguments against physicalism. It challenges Chalmers’ “Zombie argument” by asserting that a world without consciousness cannot be physically identical to ours, as the evolutionary path would have been fundamentally different. It also provides a computational perspective on Jackson’s “knowledge argument” regarding qualia, explaining how different percepts are formed from sensory data.
In conclusion, this paper offers a simpler, empirically testable hypothesis for consciousness, framing it as an evolutionary advantage that emerged from the interaction of reasoning with reactive physical substrates. It suggests that the “hard problem” of consciousness might be made easier by understanding its computational function as a mechanism to distinguish between reality and imagination, and opens new avenues for understanding and potentially implementing consciousness in artificial systems.
