40 2024.09.28 2024.09.29 2024.10.01 article M. Ljubičić (Amenoum)108. brigade ZNG 43, 35252 Sibinj, Croatia (amenoum.org)mljubicic99{EAT}gmail.com Why AI cannot be conscious. computer science ai, machines, consciousness, sentience https://doi.org/10.5281/zenodo.13854315 /authors/Amenoum.html#credits 1 Why AI cannot be conscious: The requirements for consciousness Abstract A discussion on consciousness. The paper provides possible answers to questions about consciousness, mainly what it is and what are the requirements for its emergence. Based on that, the paper concludes that manifestation of real consciousness in artificial systems is unrealistic. Introduction What is consciousness and how does it work? Where does it really come from? Is it emergent or not? So many questions, not many satisfying answers. Humanity always seems to trivialize phenomena it does not know much about. Knowledge can be exploited to serve certain interests. But lack of knowledge can be exploited as well. And humanity is not lacking in the lack of knowledge about consciousness. The beliefs about it are thriving with very questionable assumptions taken for granted. It is conventionally assumed to emerge from increasing complexity of neural networks. It is then assumed that these neural networks don't have to be biological, etc. With a chain-reaction in presumptive reductionism, one eventually comes to the conclusion that consciousness can emerge from an algorithm. How scientific is that conclusion? Is it based on reality or fantasy? Here, this will be discussed. And more. Artificial problems and natural solutions Computer science is mathematics. Nothing apart from the thing it is coded to do can emerge from an algorithm. If one wants consciousness out of an AI, one would have to code it in. And it is impossible to code it in because it is obviously not computable (if anyone believes it is, I challenge her/him to provide the code). Now, wired computer transistors are a physical thing but they're wired the same no matter what algorithm they're running. Of course, the brain is also wired the same no matter what algorithm it runs, however, the brain is conscious regardless of the complexity of that algorithm. Thus, if an machine could be conscious, it would already be conscious by running an algorithm that simply adds 2 and 2. In other words, complexity of an algorithm has nothing to do with consciousness, or its emergence. If an algorithm cannot be conscious, can the machine running it be conscious? Well, it appears, by using conventional transistors, logic, and wiring, it cannot be - regardless of the complexity of wiring, size and number of transistors. Deterministic machines in general probably cannot be conscious on their own. Something is missing. Nondeterminism on its own is unlikely to help either. What is needed probably is a coupling of deterministic and quantum phenomena. However, obviously, one cannot just couple a quantum computer with a deterministic one and expect a conscious system. If increasing complexity of a deterministic machine doesn't lead to emergence of consciousness - at least not on its own, would increasing complexity of a quantum computer do? Based on what can be inferred from biological systems, it seems that consciousness is not a binary state - amount of consciousness varies across species. It is probably proportional to the strength, or complexity, of coupling between the deterministic and quantum phenomena. And strength and complexity of coupling probably correlates with complexity of both. The key to consciousness is in the coupling/entanglement. But a lot of questions remain about the nature and mechanics of that coupling. Humans have already coupled deterministic machines with quantum machines but there is no sign of consciousness or self-awareness. It is either not there, or its amount is insignificant. Another possibility is that the consciousness is there but it is simply introverted, with no means to be expressed externally, in order for us to detect it. If it is there, how does one make it use the deterministic machine to express itself? One probably cannot easily achieve that. Such coupling (which, in biological systems, may be interpreted as part of symbiosis) needs to evolve - it cannot be simply turned on. And it really doesn't help that nothing is known about the quantum part in the coupling associated with biological consciousness. Many still assume there is no quantum part, expecting for AI to become conscious some time in the future. Some even claim it is conscious already. It is not. Without knowing true nature of consciousness, any coupling is guesswork. Best bet would be to emulate a biological system. But how even to start? We do know a lot about brains but a lot we still do not know, and the quantum part remains total guesswork. Can one evolve conscious coupling in a dish, using synthetic materials? Well, certainly not with the amount of guesswork still present. But what is the difference between synthetic and biologic? In nature, the coupling has evolved from simple molecules and chemistry, which evolve from more elementary particles. That's probably where one needs to start to evolve more complex conscious coupling. But can this evolve into something one would not consider biological? Unlikely. Why? Because nature does not discriminate between biotic and abiotic phenomena the same way humans do. In other words, there is no branching in evolution toward more complex conscious abiotic systems on one side and toward more complex conscious biotic systems on the other, there is transformation of coupling from introverted expression toward extroverted expression of consciousness. Humans are the biased ones, interpreting systems as non-living and non-conscious due to absence of extroverted expression of consciousness. Apart from conscious interaction, animals perform non-conscious interactions which are usually interpreted as subconscious and uncontrollable. Therefore, low amount of consciousness is synonymous with deeply subconscious interaction or low amount of externally expressed self-awareness. It is absurd to expect that conscious coupling suddenly appears out of nothing at some point during evolution. Rather, it is elevated from one vertical level to another. Deeper levels become elevated to higher levels, which humans may then interpret as increased self-awareness. Note that this is all relative. From human perspective, any organism operating at the same or similar level as humans do, may be considered conscious, while electro-magnetic interactions may be considered as deeply subconscious interactions, so much so that any self-awareness is infinitesimal. But is the human level the peak level of consciousness? From some higher level, human interactions may be considered as deterministic as electro-magnetic interactions are deterministic from our perspective, and humans may be interpreted as non self-aware. We may even be manipulated by some higher level of consciousness just like we manipulate atoms and molecules to build machines. While we do know how to manipulate electro-magnetic interactions we do not know how to elevate the level of consciousness of the collective of such interactions into something more, otherwise we would have created [what we usually interpret as] life already. A [relatively] self-organized collection of atoms into a molecule is something more than the sum of its parts. Apart from electro-magnetic interactions, this molecule can interact chemically with other molecules. One could say that chemistry is just a sum or superposition of more elementary interactions (i.e. electro-magnetic) but that is just an assumption. The knowledge of, so called, elementary particles is incomplete. And even with what is known, no one is modelling chemical interactions as interactions between elementary particles rather as interactions between emergent properties of molecules. Why? Because of the rapid rise in complexity, hard to solve computationally even by using approximations/shortcuts. Most likely, emergence of a molecule is also the emergence of an additional level of consciousness (or, elevation of existing levels). It is conventionally assumed that physical laws are invariant to scale. There is plenty of reason to believe that this scale invariance is relative, not absolute. In fact, I have provided enough logic and evidence for the existence of discrete vertical levels of relative invariance in my works. Molecules can be further organized into proteins, proteins into cells, etc. Again, simulation of interactions doesn't start at some lower level for the same reasons. If each of these levels represents a particular level of consciousness, which is more than the sum of individual micro-conscious entities, then it should have a physical manifestation that is more than simply the sum of lower levels. So what is it? Well, it can be interpreted as something that evolves from lower levels or something that is entangled with lower levels. In any case, it can be represented as a superposition of levels. This superposition of the collective then encompasses all other levels of consciousness, albeit perhaps not at the same time. This superposition is then likely manifested as quantum superposition of particular scale. One could now argue that quantum superposition is unstable at room temperature. Yes, quantum superposition of atoms is unstable, but superposition of atoms is not required here, rather the superposition of parts that make the atoms more than the sum of their parts. Temperature is scale relative. There are particles whose temperature is invariant to the temperature (kinetic energy) of atoms. Consciousness on each level (scale) thus represents a quantum superposition of certain scale. Note that is is not necessary for this superposition to be continuously present or infinitely stable. Stable conscious coupling may be a manifestation of periodic establishment of superposition. Consciousness likely has a frequency. In fact, this is probably required for external expression of consciousness (as it can be correlated with brainwaves). The coupling does not have to be stable either. A flock of birds is more than the sum of its parts but only briefly. A biofilm (a collective of microbes) may be a more stable coupling but still apparently not stable enough to be interpreted as an organism by humans. It may, however, be interpreted as an unstable organism, or a proto-organism). Evolution is thus required to transform unstable into stable couplings. Conclusion The phenomenon of consciousness is far from being well understood. So far it has been unequivocally manifested only in complex biological systems. Consciousness is not computable. Emergent phenomena in nature have never been described by more elementary phenomena without resorting to approximations and shortcuts. Yet, all emergent phenomena are assumed to be nothing more than the sum of their parts. The knowledge of, so called, elementary phenomena is incomplete. It is also assumed that known elementary particles are absolutely elementary. Yet, it is well known that this absolutism has always been correlated with limits in observation. All things considered, as the analysis shows, claims of possibility of any kind of elevated real consciousness in organized artificial systems has no basis in science. In other words, such claims are extraordinary and require extraordinary evidence. Some could argue that some claims in this paper are extraordinary and require extraordinary evidence as well. But, from an unbiased perspective, what is more likely? That all assumptions on emergent phenomena are true, even though they are unprovable? Or that we are not almighty gods and thus should not automatically assume that [what we consider is] the simplest explanation is correct? All assumptions stemming from overly reductionistic approaches have been proven false in the past. Isn't it time to accept the fact that nature is holistic?