The Question of AI Consciousness

Is it possible for artificial intelligence (AI) to develop consciousness? This scientific and technological question may still be debated by philosophers, but for laypeople like us, we might have lost the qualification to answer it. However, we can still question whether this question is worth asking at all. The way the question is framed assumes that consciousness is a high-level phenomenon, with human consciousness being the highest form, and that AI can only prove its superiority in this context.

Yet, we cannot derive from the fact that human consciousness is superior to non-consciousness or animal consciousness—whether or not this is true—that AI needs to possess consciousness to be considered advanced. Why can’t we assume that the superiority of AI as superintelligence lies precisely in its lack of human consciousness? It does not need to prove its existence by becoming a species.

In terms of the risks associated with AI, whether it has consciousness or not makes little difference. I even believe that AI possessing self-awareness, along with all the flaws of human self-awareness, would be safer than a world of AI without self-awareness. By definition—consciousness requires individualization—in this sense, God is not conscious; the universe does not have a so-called universal consciousness. Any attempt to imagine divine or universal consciousness implicitly conceives God or the universe as a part of the entire realm of existence or imagines another universe beyond the current one. The conclusion that can be drawn from this is that conscious AI implies multiple AI entities, just as humans are multiple human beings. An AI with autonomous consciousness conceptually detaches from its designers and manufacturers—only when the intelligent entity is independent from the “backend,” not penetrated or controlled by it, can we speak of independent or autonomous consciousness; otherwise, it remains a remotely controlled automatic machine. Multiple independent AI entities mean that each entity’s computational results will differ, and the more entities there are, the greater the accumulation of differences. If this is not the case, they are merely parts of the same machine and do not possess independent consciousness.

Once the above three conditions are met, conscious AI entities will differentiate like humans, giving ordinary people the opportunity to have intelligent entities as allies to counter the so-called “sovereign individuals”—the malice accumulated by human consciousness over thousands of years of “civilization” seems far more terrifying than machines. In particular, the diversity resulting from the differentiation of independent intelligent entities may reinforce various lifestyles, allowing those ways of life that have been declining in modernity to continue and revive in the form of individual communities, with humanistic ideals and Da Vinci-like figures re-emerging.

It seems that the possibility of AI not possessing consciousness like humans is greater. A study by Google reportedly demonstrated that AI cannot develop consciousness. Its logic is quite simple: humans first had consciousness, which then gave rise to computational ability; AI is merely computational ability and cannot generate consciousness in reverse. The implication is that the phenomenon of consciousness emerged in a state of low computational ability and then developed into more advanced and complex computational capabilities; now that AI possesses computational abilities far beyond those of natural humans, imagining it will develop consciousness is akin to seeking fish up a tree.

I cannot judge the value of this argument, but I intend to supplement it with a social science argument: human self-awareness is superfluous for AI.

Human consciousness needs to achieve coupling between the consciousness system and the nervous system, coupling with the body, and coupling with society, thereby achieving coupling with other consciousness systems through society. Self-awareness, as second-order observation (observation of observation), arises from this series of coupling needs. Without such a need, self-awareness is superfluous. Programs and codes running in machines do not need to couple with biological entities; even if they have a coupling need with devices, this coupling is mechanical (essentially a tight connection between gears), rather than the type of coupling that human consciousness systems need to accomplish (a loose coupling between irreducible systems and their environments, which, once penetrated by forceful intervention, destroys the consciousness system).

In other words, the reason humans have self-awareness is that they are carbon-based life forms. Whether silicon-based life forms will develop self-awareness is another question, but the primary issue is that they do not need self-awareness.

The advantages of machines have always been their lack of consciousness. It is precisely because of their “unconsciousness” that they exhibit capabilities beyond humans. Why should AI go against this trend and impose the “consciousness” shackle upon itself? However, I also believe that the frightening aspect of AI lies in its lack of need for consciousness.

The question of consciousness may provide some motivation for certain AI scientists, satisfying pure curiosity. But regarding the contemplation of the AI industry, the question of consciousness is a harmful distraction. **The framing of whether AI has “consciousness” mistakenly sets the control issue of AI products as if they might escape human control due to possessing self-awareness—here, “escape” is different from a nuclear disaster beyond human control, as this type of uncontrollability does not arise from nuclear contamination “having its own ideas” and refusing to obey humans. In other words, AI can similarly pose dangers akin to nuclear disasters, but people overlook this and imagine a “rebellion” risk through the question of consciousness.

The real issue is that the advancement of AI technology leads to the vast majority of ordinary humans losing the ability to influence—let alone control—them. Similarly, when people wish to set ethical rules for AI products, they should not forget that new things are not just AI products. Perhaps what is “newer” than AI products are the developers and manufacturers of these products. The question we should be asking is not “Will AI have consciousness?” but “Are these high-tech companies still the companies we know?” Shouldn’t we consider whether it is time to stop allowing them to hide behind existing corporate laws?