Beyond algorithm-based AI concepts: Consciousness and its generation on machines

Everything is simulation

20AI today relies on two pillars of intelligence: reasoning and learning. The PSSH hypothesis of Simon and Newell (1976), the basic hypothesis of AI, states that intelligence could be simulated by mechanical symbol manipulation (e.g. on a computer). This has not really been contradicted to date. Learning and thinking does seem to be mechanizable (“mathematizable”); all AI relies on mathematical methods. However, this is a misconception. Because only this part of thinking which can be mechanized has been mechanized so far. And everything that can be mechanized can run on a computer, therefore it looks as if a computer can think. But this is only a linguistic convention: One says that the computer can think, because it implements the mathematical part of thinking so well, that a nonequivalence of simulation of thinking with thinking would look like splitting hairs. But human thinking goes beyond logically correct thinking. In particular, humans can arbitrarily switch between logics, deal with contradictions and even with circular reasoning. And so, one quickly sees the limits of the purely mathematical method of thinking. All current AI systems fail sooner or later due to the complexity limit of second-order predicate logic.

Limits of logic and logical thinking

21Since the work of Kurt Gödel (1931) it is known that 2nd order predicate logic can lead to statements whose truth is not computable. However, algorithmic computer systems with non-verifiable truth pose a serious logic problem. A similar logic problem applies to circular reasoning. “All Cretans lie,” says the Cretan. But is he lying? For a computer, this is a situation where it can't come to an end with the calculation. A human being, however, could say, “yes, the Cretan there is lying, I have already met him in Athens”. A human solves the problem by getting out of the logic system, and getting back in, any number of times.

22As already pointed out, humans can deal with unknown and even contradictory truths and also with any number of circular reasoning. The reason is probably that humans have not experienced their own I as an infinite regress; humans have been able to cancel the circular reasoning of their own I. Their ME contains themselves completely, because the ME originates from the brain and yet contains the whole body from which it originates. Systems that master this logical difficulty can probably deal with all logical antinomies, but this is only a thesis. Another reason might be that humans do not “live” in logics but have invented logics as a symbolic form to structure experience in a formal way.

23In any case, computers do not contain themselves. One could use AI programming languages to develop software which contains itself, but from this no experience for the system arises, it would remain purely syntactical. Computers permanently reach their “thinking limits”, humans not at all. That computer systems beat humans in certain mental tasks like GO or chess is simply because this part of thinking could be fully algorithmized, which is why computers outperform humans in these areas. As soon as something is algorithmizable, a computer will solve the tasks better than a human, but infinitely many problems are not algorithmizable, as every mathematician knows. Thus, the number of all possible algorithms on all computers of the world is (only) countable (infinite), but the variety of all problems is over-countable (infinite). Algorithmization as a problem-solving method is thus a good but severely limited tool for realistic problem solving that nature presents us with. However, today's AI can only turn to algorithmizable problem solutions; the rest is in principle unattainable.

Limits of learning

24Machine learning systems require data to learn their models. Thus, in principle, these inductive learning methods cannot generate general knowledge. For example, if an ornithologist has built a table in which the plumage of all swans is white, the AI – which is supposed to generate rules from this – must conclude that “all swans are white”. But that's not true, as any human knows. Machine learning is a statistical concept, which usually does not lead to universally valid knowledge.

25A human being, however, can use the results of a learning and observation process in a creative act to generate completely new and valid knowledge (such as the law of gravitation by observing falling apples). Now a creative act is characterized precisely by the fact that no algorithm exists that specifies how to get from a starting point A (e.g. observing falling apples) to a new creative state B, thus algorithmic AI computers cannot perform any creative performance. They can apply learned knowledge in the extrapolation space, but any expert knows that beyond a certain distance from the learned supporting points in the learning data space, only wrong results are produced. AI systems are not creative and they cannot develop general-purpose rules from learning results. Humans can do that very well.

Limits of simulation

26Although the mathematical-algorithmic method reaches its limits as mentioned above, nevertheless mathematical methods for the description of consciousness are already recognizable in rudiments. So, one could assume that consciousness like thinking and learning could be generated (simulated) mathematically on a computer. But this is a false conclusion. Just with consciousness, especially with perception as the lowest level of consciousness, the limit of any simulation becomes apparent. If one has developed mathematical equations for the big bang and runs them on a computer, their application in the processor does not cause any physical changes in its environment. The simulation is said to have no “causal power”, so it cannot interfere with the physics of the environment – except for processor heat and such like. And so it is with all mathematical simulations. A simulation of chemical equations does not produce chemistry; a simulation of biological processes does not “live”. Mathematical descriptions are essential, but in no scientific discipline has it occurred to anyone that if you run the equations in the form of software on a computer, then more than simulation is happening. But AI suggests otherwise. Because it equated thinking with the mathematics of thinking and learning with the mathematics of learning, and was very successful in doing so, while it denied the difference between real process and simulation and postulated that the AI machines were actually intelligent.

27However, the success of AI stems from the fact that the rational thinking and learning of humans has been precisely modeled mathematically in advance and can thus of course be run on a computer. This is a completely normal process. When a computer adds two numbers, it is calculating. Or does it only simulate calculating? For philosophers this might have been an important question, but for engineers, computer scientists and users of AI it was not. If the simulation is sufficiently exact (means enough for selected technical applications), one does not distinguish it linguistically anymore. One says that the computer thinks, learns and calculates. But this is just a linguistic convention, because it should be clear that the distinction between something that happens and the statement that something happens is not only important, but fundamental.

28This distinction becomes particularly clear when we are no longer modeling mathematical processes (and exactly those are rational thinking, learning and calculating), but physical, chemical or biological ones. Here, of course, the simulation does not reach the “causal power” of the original process. Thinking and learning can be “generated” mathematically as explained above (at least a large part of it), the crucial thing now is that one cannot do that with consciousness, since consciousness is at least a physical property. Also weight, size or form cannot be generated mathematically, but these physical properties can only be described mathematically.

29One could describe the running of mathematical equations on a computer as a physical process, because finally register entries are read out, written in etc., but this physical process in the memories and circuits is not specific enough to produce physically “more than just waste heat”. The digital computer does not physically distinguish what kind of equations are running in the processor, this is also the reason why intelligence and consciousness cannot correlate in a digital computer, as mentioned earlier. From this example you can also see that there is no emergent behavior of a computer. The running logics on a computer do not emerge from the underlying physical processes; emergence might turn out to be the biggest scientific illusion later on.

Example 1: Perception cannot be simulated

30A concrete example where the limitations of simulation are even more apparent than in thinking and learning is perception.

31For the last 10 years, AI in the field of machine vision has been racing from one success to the next, Deep Learning Networks, for example as CNN, have established downright revolutionary possibilities in image processing. While classical backpropagation networks can already learn anything that is learnable (and continuous), they have failed in image processing because the neural image evaluations were not translation and rotation invariant. Convolutional neural networks (CNN) have been used to overcome these difficulties.

32And yet, it is in machine vision that one recognizes the great difficulties of AI. To put it in a nutshell: AI systems of machine vision cannot see, they are blind to their external environment. The reason is not really surprising; today's AI systems simulate seeing. but seeing is not a mathematical process, not at all, but a thoroughly physical one. This, of course, has enormous consequences. To distinguish between dogs or cats, an AI needs tens of thousands of images, but a child needs at most 10 or 20 different cat and dog images. The difference in performance between humans and machines is so dramatic that image processing will have to be done differently in the future and mathematical deep learning systems will have to be fundamentally changed. But how?

33The perception of the human being reveals surprising things when looking closer. The human being sees, as everybody knows from own experience, the images where the photons “start their way” and not where they “arrive” (on the retina). The term “seeing” (better: visual perception) therefore gets a clear physical meaning. Seeing is the perception of electromagnetic waves from the place of their starting point. It is as if the photon on its way to the eye carries its place of origin at the start. When a photon hits the retina, one should expect that there (only) color and intensity are recorded. But the starting point of the photon? How does the starting point arrive at the retina? This can only be explained quantum mechanically. Of course, using optical laws, the starting point P(x,y,z) in Fig. 2 can be back-projected from the point of impact on the retina P'(x',y',z') and the refractive properties of the eye lens, but why do we "see" the starting point P outside the brain? How does the visual re-projection from P'' (in the visual cortex) to P happen (Fig. 2)?

Fig 2: Seeing a point P in the outside world

34If one electrically excites the retinal cell P'(x',y',z'), then no bright spot P(x,y,z) appears for the affected person in front of his head in the external world. The author posits that if one were to initiate - directly behind the retinal cells - exactly the identical electrical action potentials A(x',y',z') into the visual cortex that would normally be evoked by the retinal cell during vision, then the subject will still not see an object dot P(x,y,z) of the environment. Vision, and also hearing, cannot be derived solely from the electrically evoked nerve potentials A that are conducted from the sensory organs to the brain. A further processing step must be introduced that leads to an image P''' in consciousness. This approach is not state of the art, but it is a compelling necessity from the author's point of view.

35When a human sees something, it is not a model formation in the visual cortex which is separate and separable from the original. No, the person is (from the point of view of the so-called First person) visually "merged" with the viewed object (P''' ≡ P), a technically considered "incredible" process. Dreyfus (1985) says that people do not represent their environment, they are in it. From the sighted person's point of view (First person), model in consciousness (P''') and original (P) cannot be separated at all. Such a separation can be learned, but for the time being perception (external) and object (external) are identical. It is different from the point of the Third person. For a researcher analyzing a sighted person during vision, there is the original object (P) in the environment and the neural image (P'') in the cortex (which he can detect by MRI, for example). But from the point of view of the First person (the sighted person himself) the model P'' in the cortex does not exist, because he does not see the model in his head, he sees the object P''' in consciousness and therefore outside in the environment. Humans (and certainly also highly developed animals) therefore have two ways of signal processing, an internal representation of the environment in their internal neuronal networks (P'') and additionally a perception of the object in the outside world in the consciousness (P'''). Computers only have an internal representation of the environment, they do not see out of their cameras into the world. As strange as it sounds, humans can do this.

36But how exactly does P''' get back outside to P? This is a completely new research question, which precisely supports the author's thesis that consciousness is needed to see, especially consciousness with hypercomplex energy states. This effect is the essence of the approach presented.

37Seeing and thinking are two completely different mental processes. While thinking can be easily approximated mathematically, seeing cannot. That is the difference. And in computers, the process of seeing is quite different from that in humans. Computers reconstruct a mathematical model of the environment from received electromagnetic waves and recoded electrical signals in the CCD chips of their cameras, which we also mistakenly call seeing because electromagnetic waves are used for data acquisition of the environment. But one could just as easily use acoustic waves as the basis for model building (of course, their resolution is lower); in that case everyone would understand that this form of data processing does not lead to vision. All methods of deep learning for machine vision are now pure mathematical simulations, with all the resulting disadvantages. For the computers of today's machine vision, the outer boundary of their world is the chip of the camera, there is no further “outside”, it is an impassable, because physical hurdle for it. The consequence of this is highly explosive for many applications, such as autonomous driving: computers see nothing, their tables of zeros-and-ones are exhausting mathematical reconstructions of an outside world unreachable for “them”. With no deep learning system in the world, it is possible to “project” the images from computer memories back into the outside world, but that is exactly what humans do (Otte 2019).

38In summary, software vision does not exist. Software seeing is a principled delusion to which researchers are subjected and a serious limitation of any software AI.

Example 2: Semantics cannot be calculated

39Let us come to another example of limits of simulation. Computers of today's design cannot comprehend the semantics behind the statistical syntax. Of course, pseudo-semantics can be derived from context, otherwise there would be no language translation systems or state-of-the-art speech engines, for example, but semantics is a matter of recognizing the meaning of things independently. When an AI sees a tree, it creates a table of zeros and ones as a representation inside it, as just described. This table is not linked to anything in the original environment. But meaning is created precisely through linkage and association. Now, one can program any number of associations, knowledge-based systems provide the basis for a multitude of such links. But every single association must be foreseen. And meaning goes far beyond associations. In Otte (2016) it is stated that meaning always arises at the receiver of a message and cannot be contained in the message itself after all.

40Let us consider, for example, a language-processing system that processes numerous texts a,b,c,... one after the other and arrives at some point at text k, in which it says: “The recipient of text k will be destroyed after reading”. An AI system of today's design will process this text k in the same way as the texts a, b, c. If the syntactic information quantity H of a,b,c and k is the same (H_a=H_b=H_c=H_k), an observer who monitors all electrical voltages and currents of the system will still not be able to measure that the message k differs semantically from the messages a,b,c, because the receiver does not understand the message, and syntactically this message contains the information quantity H_k, with H_a=H_b=H_c=H_k.

41If one would assume at all that the meaning of a message is (only) recognizable in the electrical changes of the currents and voltages of the receiver, then for example two messages a and b with different syntactic amount of information H_a (“Hello”) and H_b (“Hello world!”) would have per se also different semantics, because their processing at the receiver would lead necessarily to different currents and voltages. But even this interpretation can be wrong. Even two messages i and j that mean nothing to the system, for example (whose semantic information set is therefore zero for the systems), would lead to different processing steps if message j is, for example, a hundred times longer than message i. Therefore, encoding the semantics alone in the measurable processing sequences of the receiver is not possible at all. But semantics cannot be encoded in the message itself either, since its meaning lies mainly in the receiver's reaction to that message. So where is semantics encoded?

42If one considers semantics as mathematically “orthogonal” to syntax on a metaphoric level, then one recognizes that systems with consciousness (whose consciousness processes could be also mathematically “orthogonal” to the neural processes, Fig. 1) are predestined to understand semantics. Semantics is recognized in consciousness, and only there. It is assumed that the semantic information quantity is encoded in the change of the imaginary state of a system, in this case of the receiver system. (The term “imaginary state” was introduced at the beginning, describing physical phenomena with immaterial energy levels; this will be elaborated in paragraph 3). Therefore, conventional AI computers cannot evaluate any semantics; their imaginary states (in the digital circuits) are completely independent of the processing tasks.

43If a classical AI syntactically learns that rain is wet, it means nothing to an AI because “being wet” is a subjective experience in consciousness. An AI without consciousness can attach many actions to the measured state of “being wet”, such as “sidestep”, “send error message”, etc., but it has no subjective meaning because this has been defined only for consciousness. One could program human semantics into these systems, in the sense that the state "to be wet" is to be avoided, because otherwise the system rusts or the circuits become defective, or the car can slip. But this is still not a subjective experience of the system itself, and “being wet” still means nothing to the system.

Consciousness has unusual physical properties

46Currently there is no chance to understand human consciousness mathematically or even to model it. Only a massive reduction of the complexity of the problem allows an engineer to deal with this topic at all. Of all the manifold properties of consciousness, only the physical ones will be highlighted in the following. If the physical properties of human consciousness are known, then there is hope to establish processes on physical systems that express consciousness processes.

47We thus consider human consciousness in its greatest possible abstraction as a physical process. But consciousness processes have extremely unusual physical properties:

1. Consciousness processes are not physically measurable, because up to now it has not been possible to measure consciousness directly, despite all advanced measuring methods. Consciousness is an inner state, an experience from a very private point of view (so called Qualia), without any access from outside.
2. Consciousness processes can store information.
3. Consciousness processes have a temporal behavior, i.e., they change their information state over time.
4. Consciousness processes are non-spatial, one cannot assign x,y,z coordinates to thoughts or dream images.

49Even if consciousness processes, better: their contents, cannot be measured directly, their effects in the brain must of course be measurable. Consciousness processes are just not understood as epiphenomena, but they have permanent effects on the neuromorphic structures of the brain.

50With these physical results we come to the discussion of important scientific relations. In the following it will be explained mathematically why consciousness processes were introduced at the beginning of the essay with the term “imaginary state”.

Mathematical description of physical aspects of consciousness processes

52It is not that mathematics prescribes the nature of things, but in the history of natural science it has been shown that from time to time problems occurred which could be solved adequately only by “new” mathematical approaches. Just as real (e.g. transcendental) numbers – which already arise from the solution of simple geometrical tasks – could give us some more understanding about “infinity”, the above introduced “imaginary numbers” could make us aware of a physics (that of quantum physics) which has unusual effects. The “mental phenomena” to be described are now even more unusual than quantum phenomena and it suggests itself that new mathematical numbers are indeed needed to describe these phenomena adequately. Hypercomplex numbers could point to physical processes whose energy contents are classically not measurable. Hypercomplex (resp. mental) phenomena are thereby not separated from the real objects of nature, but would be intrinsic properties of certain objects, which produce just these properties. Thus, the mathematical description is ultimately about showing that "mental phenomena" fit into the present scientific world view, but extend it fundamentally.

53The development of natural sciences was usually based on mathematical descriptions. However, as stated above, mathematics itself must be adapted from time to time. While in the first times of the natural sciences the algebra of the natural, whole, rational and later of the real numbers was sufficient for the description of nature, quantum physics has been described on the basis of the so-called Complex Algebra since about 100 years. Complex numbers consist of real and (the already introduced) imaginary numbers, the last ones are the eponym for the “imaginary” of states of consciousness emphasized in the essay (see insertion).

54Until recently, real and complex numbers had sufficed to describe all natural phenomena mathematically. But the physical analysis of the phenomenon “consciousness” suggests something fundamentally new, namely that “mental oscillation and wave phenomena” could be immaterial (imaginary). But not only that. Even the oscillations and waves of quantum mechanics already possess a kind of imaginary (they are usually so-called “possibility waves”), otherwise one could not understand many effects, as for example the measurement problem. But their imaginary is such that after a measuring process the received mathematical values represent real "probability amplitudes" for the occurrence of physical values. This is not so with consciousness processes. Consciousness remains physically immaterial even after measurement (the process of becoming conscious). Mathematically expressed, the calculation of the energy of mental processes must therefore lead to imaginary values, because (only) this is a good explanation why mental processes are in principle not measurable with conventional measurement techniques. There’s a bunch of possible explanations. One being that the experimental setting used to measure consciousness could be ill-constructed, another being that the conceptual understanding of what consciousness is could be wrong. Or another explanation would be that consciousness phenomena do not exist at all. But this is obviously wrong. Therefore, the generation of consciousness in technology is advantageous, because it is predicted that technical systems with and without consciousness will be seriously differ from each other, which will be clearly measurable).

55Now, the physical properties of consciousness demanded above (1.-4.) can be modeled neither with real nor with complex numbers. Only one hypercomplex algebra can describe the physical problem of imaginary mathematically correctly, since such an algebra has several imaginary quantities, Fig 3. From initial considerations, the hypercomplex Quaternions seem to offer themselves for modeling mental processes (see Insertion and Otte 2011). However, such an attempt fails, since e.g. the necessary oscillatory phenomena cannot be formulated with it; that's why an algebra deviating from Quaternions has been proposed for describing "mental processes", the so-called Bi-Complex Algebra (Hertig 2014).

The Algebra of Bi-Complex numbers

56Particularly interesting with hypercomplex numbers is how the imaginary numbers can be multiplied (and divided) with each other, which determines many of their later mathematical properties. Fig. 3 shows the multiplication table for the imaginary numbers i, j and k, which is the basis for formulating oscillation phenomena and Fourier transformations for imaginary waves. Thus, there is (since 2014) a 4-D algebra based on hypercomplex numbers <1,i,j,k> which admits all physical needs, in particular to store information on imaginary wave functions.

57Numbers q from this Bi-Complex Algebra (B) consists of one real number and three imaginary numbers i, j and k, here is an example of such a number: q=5+2·i-3·j+7·k.

58Philosophically, it might be interesting that there is only one hypercomplex 4-D algebra that satisfies all the necessary properties. For example, the multiplication “i·j” must result in “-k” (Fig. 3), anything else will not lead to success, and so it is with all the other multiplications given. Thus, the results shown are inevitable.

Fig 3: Mathematical fundamentals of consciousness: Bi-Complex Algebra (1/2)

59The importance of the algebra B is to be emphasized. The formulation of oscillation phenomena is necessary because oscillations are always used when information is to be stored by modulation; the whole communication technology is based on the idea of information storage on oscillations and waves. And, of course, the storage ability of information in consciousness is a very important property, without this ability one would not need consciousness at all. If one now extends - as suggested in the paper - the idea of information storage to phenomena whose energy levels are hypercomplex, the algebra B could provide the basis for describing “mental phenomena”. (An exact mathematical analysis of the algebra B is given in Hertig 2014).

Fig 4: Mathematical fundamentals of consciousness (2/2)

60In this paper it has been repeatedly said that mind and consciousness are imaginary states of a system. If one follows this argumentation, however, the physical effect of consciousness processes on material processes seems (still) relatively unclear, since it obviously threatens to violate the causality of physics, if non-energetic phenomena can interfere with material physics. However, if such an effect happened, for example, via the change of quantum processes (ψ(x)) in the microscopic processes of a system, causality would be preserved; more detailed explanations are given in Otte 2016. In Fig. 4, equation 4.7., one can already see how a multiplicative interaction of imagined quantum mechanical waves within brain tissue ψ(x) with imaginary waves of consciousness ɸ(y) could be formulated. The result is explosive: Immaterial processes, i.e., also phenomena of consciousness, could act into the material processes even without real energy. Of course, the result was tacitly expected, because if it were otherwise, consciousness would have no biological sense at all and in technology one would not need machine consciousness. One measures consciousness by its effects, because consciousness without effects into the material world we would not need at all.

61With the above representations the difference of consciousness phenomena to quantum physics can be clarified once more in its different mathematical description: Quantum physics describes its phenomena with complex numbers z = a+b·i. Possible imaginary phenomena are thus expressed in complex-valued terms. In contrast, the phenomena of consciousness are described with the imaginary numbers j and k. From a mathematical point of view, the imaginary of quantum physics is of a different nature than the imaginary of consciousness (Fig 5).

Fig 5: Hypercomplex wave functions for the description of mental states

Summarizing the mathematical paragraph about immaterial processes

62Consciousness processes are immaterial phenomena (with mathematically imaginary energy levels), which arise on suitable physical objects. In the human brain such consciousness processes form by themselves, which cannot be measured from the outside (to repeat: because of their imaginary energy level). Immaterial processes have thereby special physical properties, which the human brain uses for example for the process of seeing, since the properties make the (apparent) outside projection of images possible, as discussed above. The inner perception (Qualia) of human beings - based on those immaterial processes - can be described mathematically in its basic features. Using the mathematical description, technical specifications can be obtained.

Neuromorphic computers as a technical basis for generating machine consciousness

63From the mathematical description above it is clear that consciousness processes are associated with quantum processes (at least, however, with electromagnetic photon processes), but they are not identical with them. The thesis propagated by Penrose and many others (Hameroff 1996; Stapp 2004; Penrose 2017) - that quantum processes in the brain are responsible for the expression of consciousness – has to be extended.

64This extension is mathematically complicated (as we have seen), but nevertheless it shall be repeated here again in a scientifically correct formulation: The application of energy operators to quantum processes leads to mathematically real-valued results. The application of energy operators to the imaginary wave functions used in Fig. 4 leads to mathematically imaginary results in base k (see Hertig 2014 and Otte 2016). These different mathematical forms of description are necessary because fundamentally different physical phenomena are described: Quantum processes take on real physical values through measurement processes (with a certain probability), consciousness processes remain unmeasurable even after becoming conscious, in the mathematical description this is implemented by the hypercomplex representation.

65Ultimately, consciousness processes are a kind of “hyperquantum process” because their physical properties are “even more imaginary” than those of quantum processes themselves. However, quantum processes (at least electromagnetic processes) operate as an interface of information exchange between immaterial and material processes (Fig. 6).

Fig 6: Mind and mater interaction

66It is obvious, of course, that consciousness processes will only play a role in technology if they – despite their special imaginary nature – can affect technical systems. In Fig. 4 simplified interactions between wave functions to base i and j, k have already been shown (e.g. equation 4.7). Such interactions could represent the physical basis for the effect of immaterial processes on material processes. Consciousness processes, according to this interpretation, would never act directly on matter, but would only ever change the probability amplitudes of quantum processes, which in turn could then lead to altered physical measurements. Systems with usable consciousness must therefore contain quantum processes; this mathematical insight is technically relevant.

67Since quantum processes are present in every technical device, the design of machine consciousness is about using systems in which the quantum processes can be used for the processing of tasks. This is not the case with a pocket calculator or a digital computer, for example; here as well as there, the results are generated by “classical” electrical or magnetic operations in electronic circuits.

68Examples where quantum processes directly affect the results of tasks are quantum computers and neuromorphic computers. Neuromorphic computers are computers that use physical resources to get as close as possible to the neural structures of the human brain. While neural networks in a digital computer are pure mathematical constructs, neuromorphic computers are physically approximated as closely as possible to neuromorphic structures (Wan 2022 ).

69A concrete implementation example for the use of neuromorphic computers are memristors. The memristors physically replicate the synapses of neuronal structures correctly. By changing the voltage, one can increase and/or decrease the weights (resistance values) of the memristors and thus implementing the synapses of the brain in a physically correct way. Quantum processes take place in the memristors, so that they can provide the physical basis for the expression of consciousness effects. However, quantum computers and, in particular, photon computers are also most likely perfectly suited for the manifestation of consciousness effects on physical machines.Digital machines, on the other hand, are not. Neuromorphic computers thus represent a concrete hardware basis for the generation of machine consciousness.