Abstract
The fulcrum of this work is knowledge: what it is and how it is generated within the context of a capitalist society. First, Marx’s analysis of the objective labour process is extended to the mental labour process. Then, objective and mental labour processes are defined in terms of objective and mental transformations, with consideration paid to which of the two types of transformation is determinant. This requires a discussion of dialectical logic and formal logic. Within dialectical logic, two types of processes are introduced: open ended and pre-determined. It is argued that computers (both traditional and quantum) and Artificial Intelligence cannot generate new knowledge, because they (a) rely on formal logic, i.e. they cannot engage in open-ended dialectical processes, and (b) are impervious to social determination. Connectedly, Artificial Intelligence systems such as ChatGPT cannot be a substitute for human thought or writing, because of the inevitability of ‘model collapse’. Next, focus is shifted to a specific form of knowledge: the ‘Copenhagen interpretation’ of quantum mechanics. It is shown that this interpretation is steeped in irrationalism and that it is a variant of pro-capitalist ideology. Finally, the social-historical roots of this ideology are revealed.
1 Introduction
The pivot of this work is knowledge: what it is and how it is generated under capitalism.1 Its theoretical core can be found in Capital Volume I, in Marx’s analysis of the labour process, which is one of his towering theoretical achievements. Contrary to received wisdom, Marx’s analysis is not necessarily restricted to the production of physical (or better said, objective) use values. His investigation of the labour process can and should be extended to the generation of knowledge, i.e. to the mental labour process, whether it produces scientific knowledge, a religious belief, superstition, and so on.
Unfortunately, this line of research has remained underdeveloped, probably because of the misguided belief that physical production determines, and thus is more important than, mental production. There have been, though, two moments when this investigation could have been pursued but was not. The first was in Lenin’s Materialism and Empirio-criticism. The second was in the debate following the publication of Braverman’s Labour and Monopoly Capital. Both moments have been important steps forward. Lenin advanced Marxist epistemology against idealism, and Braverman sparked an in-depth analysis of the labour process. But, in both cases, the specificity of the mental labour process remained uninterrogated. The need to proceed further in such a direction is even more acute now that Artificial Intelligence (hereafter, AI), computers, quantum physics and their practical applications are increasingly ubiquitous in the economy and in society at large. These developments pose new theoretical problems which, if left unaddressed, will undermine Marxism’s relevance as a revolutionary theory of modern capitalism.
2 Human Thinking
The building block of Marxist epistemology, as submitted here, is the notion of transformation, the outcome of a process of change. There are two types of transformation. Objective transformations change objective reality, i.e. what exists outside of our perception or knowledge. This concerns natural phenomena – those which are not caused by human action (such as a natural waterfall) – or objectifications of human labour (such as a man-made waterfall). Mental transformations affect our knowledge of objective transformations as well as previous mental transformations. Knowledge never begins from a tabula rasa.
Objective transformations are material because they are the application of (material) human energy onto objective reality, which is itself material. This is uncontroversial. The question that remains is whether mental transformations, and thus knowledge, are material.2
The thesis of the materiality of knowledge is of fundamental importance for Marxist economic theory. Contemporary society is increasingly based on the production of knowledge as the outcome of a collective labour process. Marxism is destined to become irrelevant if it is unable to explain why and how the production of knowledge can also be a production of value and surplus value. The importance of this point is not simply theoretical: ‘Global R&D [research and development] expenditures tripled from $726 billion in 2000 to an estimated $2.4 trillion in 2019’.3 Theorising the increasing economic importance of the production of knowledge requires that we address the question of its nature.
This article argues that knowledge is material. It is the outcome of the cognitive process (the mental labour process) that is the expenditure of human energy, which is itself a material process. More specifically, the cognitive process is a process of change in the nervous system, in the interconnections of neurons in the human brain. These interconnections are called ‘synapses’. If synapses are material, then their effects are material too, and so must be knowledge as the outcome of such interconnections and effects. Therefore, knowledge is material even if intangible. There are no immaterial transformations and thus there is no ‘immaterial labour’, pace some ‘workerist’ authors.4
To deny the materiality of knowledge is to ignore the results of neuroscience. However, while changes in synapses make possible its generation, knowledge is not only materially determined but also socially determined. It is the perception in our mind of the myriad of social relations and processes that constitute society.
The above differs significantly from traditional Marxist epistemology, which rests on Lenin’s view. For Lenin, the sole property of matter is its objectivity, or its existence independent of our perception. This is indisputable. But, for Lenin, knowledge is not material. Rather, it is a reflection in our mind of material reality: ‘The mastery of nature manifested in human practice is a result of an objectively correct reflection within the human head of the phenomena and processes of nature’.5 So, on the one hand, Lenin links knowledge to objective reality and thus rejects idealism. Yet, on the other, he bypasses the analysis of mental transformation through which objective reality is configured in our minds.6 Reflection suggests a passive and mechanistic view.7 True, it can be argued that knowledge is a reflection. Still, it is not a passive one, because it interacts with and, in doing so, changes objective reality. How, then, and through which process, can a non-material reflection of materiality interact with that materiality? Clearly, the reflection of materiality must be material itself.
Despite these lacunae, the non-materiality of knowledge became the official view of the Soviet Union, though with an important qualification. According to the Handbook of Philosophy, ‘ideas are a reflection of reality in the human consciousness … [and] are influenced in their formation by the character of the social structure’.8 The question, then, is: if knowledge is not material but a reflection of material reality, what is it? The official Soviet theory of knowledge has never been able to solve this conundrum. The Great Soviet Encyclopaedia speaks of the ‘interrelationship of material and spiritual phenomena’ (thus implying that ‘spiritual phenomena’, whatever they are, are not material) while maintaining that ‘consciousness constitutes a special property of highly organised matter’ (thus implying that consciousness, as a property of highly organised matter, is material).9
The problem is not solved by distinguishing between the material base and the essence of knowledge. This does not solve the riddle but submits a political reason for why knowledge cannot be material: if it were, the distinction, and thus the struggle, between materialism and idealism would be meaningless. This is no theoretical argument. Moreover, it is incorrect, for it is only by granting materiality to knowledge that idealism is delivered a final blow. The distinction between the material world and the world of ideas is a residue of idealism. The world of ideas, too, is material, and this is the only coherent materialist position. All that exists (both the objective and the mental) is material, and knowledge, besides being material, is socially and class determined.
How do humans generate knowledge? In other words, what is a mental labour process? As Marx masterfully shows in Capital Volume I, the objective labour process is a sequence of transformations of objective use values into new objective use values. Our task is to use the same pattern to theorise the mental labour process.
As mentioned above, transformations are both objective and mental. In objective transformations, labour power transforms the means of objective transformation (such as a hammer) and the objects of transformation (such as wood) into new objective outputs (such as a table). In mental transformations, labour power transforms its own knowledge as well as knowledge contained within objective means (such as within computers) and objects (previous knowledge) of mental transformations into new knowledge.10 The outcome of a transformation has a use value. Similar to objective use values, mental use values are the uses that can be made of knowledge.
Objective transformations necessarily imply mental transformations, because cerebral activity does not stop when objective transformations are carried out. Similarly, the generation of knowledge entails the transformation of an objective use value (for instance, the use of a pencil and paper or the use of a chair on which one is seated). Thus, all labour processes necessarily entail both types of transformation. But how can two different labour processes arise from the same transformation?
3 Dialectical Thinking
The answer to the above question requires a discussion of dialectics. As is well known, Marx did not leave us an explicit treatment of how he conceived of dialectics, even if he did say that to do so would not take more than a few pages. What is submitted here is extracted from Marx’s work, fits within the general architecture of that work and is indispensable for Marxism’s further development. The following are its basic features.11
In this section, dialectics is considered as a method of social research.12 It rests on two pillars. First, social reality is understood as always both potential and realised. Consider the labour process. Before it begins, its inputs are realised, actual phenomena. When they enter the labour process, and for its duration, they become potential elements of the output. When the labour process ends, they are constituents of the output, potentiality having become realised as/in the output. More generally, potentialities are forms of material existence (both mental and objective) in the process of transformation, i.e. in the process of being transformed into other forms of material existence. During this process, they lose their specific form and become formless potentials. At the end of the process of transformation, they are realised in a new form, as a new realised, or actual, form. Therefore, realisations are forms of material existence which have taken a specific shape as a result of a process of transformation, i.e. as the realisation of potentiality. Reality is the continuous movement or transformation of actual (i.e. realised and thus shape-full) objective and mental use values into shape-less, potential ones, and subsequently into new, actualised objective and mental use values.
The potential elements of reality are (objective and mental) material; they are elements of this world, not of some imaginary, immaterial or mystical world. There is nothing mysterious about potentiality. Potentials are elements of reality that are in the process of transformation, or in the process of taking a new specific form. This implies that the transformation of potentials into realised entities, and vice-versa, is a temporal process: inputs enter the labour process at a certain time; they are subsequently changed into potentials and finally exit the labour process as its output, in a new, realised form. This sequence also applies to the generation of knowledge. Any theorisation of reality without time belongs to science fiction rather than science.13
The second pillar is the distinction between determinant and determined elements of social reality. A labour process always requires both objective and mental transformations, but it is always either an objective or mental labour process in accordance with which type of transformation is determinant.
Consider two instances: A and B. A is said to determine B if A is the condition of the existence of B, and B is said to be determined by A if B is the condition of the reproduction or supersession of A. And so, in the objective labour process, objective transformations determine mental transformations, meaning that outputs are new objective use values. In the mental labour process, mental transformations determine the objective ones, meaning that outcomes are new knowledge. It is thus mistaken to think of objective labour – sometimes erroneously called physical or manual labour – as separated from mental activities (or mental transformations).
The use of a commodity reveals which type of transformation has been determinant in its production. If mental transformation was determinant, the commodity is used because of its knowledge content. Its production would have thus involved a mental labour process. On the contrary, if objective transformation was determinant, the commodity is used because of its objective features. Its production would have involved an objective labour process. These claims do not imply that the use of a commodity determines the nature of its production. Something must be produced before it can be used – that is, it cannot be used before it is produced. In time, its use reveals the nature of its production, and it is the general use that reveals which of the two types of transformation has been determinant. Anomalous use can diverge from the generalised use, but this does not change the nature of the labour process. Mental labourers are those whose determinant function in a labour process is to transform mental use values. They can do so from within either a mental labour process or an objective labour process. A janitor, however, is not a mental labourer, even if s/he works at an AI firm.
Besides material reward or status, some mental labourers are motivated by their curiosity. Yet, under capitalism, their curiosity is other-directed by capital. An example is the digitalisation and codification of knowledge because ‘it can be owned, transferred, and concentrated very easily’.14 These are preconditions of the privatisation and commercialisation (such as in the application of intellectual property rights) of knowledge.
Finally, writing a computer programme is a mental labour process (in which mental transformations determine the objective ones). However, if the knowledge produced by this labour is subsequently used in the construction of a bridge, it becomes an element of (and is determined by) an objective labour process (in which the order of determination is inverted). Thus, it is mistaken to hold that objective labour processes are always determinants of mental labour processes. Analogously to transformation, both types of labour processes can be determinant or determined. This is the structure of the societal labour process.
4 Dialectical Logic
We can now sketch the difference between formal logic and dialectical logic. This is needed to understand the difference between knowledge that is produced by humans and information that is produced by machines (or computers).15
Formal logic rests on three basic laws.16 The law of identity states that something is equal to itself – that is, A = A. This is a truism. It stresses what it is, not what can become. The law of the excluded middle states that A = A is either true or not true – i.e. either A = A or A ≠ A, with no third possibility. The law of non-contradiction states that two contradictory propositions cannot both be true. This is to say that A = A and its denial A ≠ A cannot both be true.
Dialectical logic is also based on three basic principles. The first is that phenomena are always both realised and potential (see above). For example, ‘the plant, the animal, every cell is at every moment of its life identical with itself and yet becoming distinct from itself ... by a sum of incessant molecular changes ... even in inorganic nature identity as such is in reality non-existent’.17 Or, to take another example, a commodity is only potentially such as long as it is not sold.18 The second principle states that realised social phenomena are always both determinant and determined (see above), depending on the context within which they are placed. The third principle stresses that realised phenomena can be either tendential or counter-tendential, given that tendencies determine their own counter-tendencies.
Let us now combine these three principles of dialectics. If we take two phenomena – A and B – a relation is dialectical if A determines B. This is to say that their relation is dialectical if A in its realised form contains B, as a formless potential, as a condition of B’s realisation19 and if B, upon its realisation, ceases to be potential and becomes the realised condition of the reproduction or supersession of A. It follows that social phenomena are subject to constant movement and change: from a realised state to a potential state, and vice-versa; from a determinant state to a determined state, and vice-versa; and from a tendential state to a counter-tendential state, and vice-versa.
The above implies a number of features of dialectical logic. First, since a potential phenomenon is different (because it is formless) from its realised form, a phenomenon is the unity of identity and difference. As a realised phenomenon, it is identical to itself; but as a potential phenomenon, it is also different from itself because it is different from its realised form. It is only by considering the realm of potentialities that the otherwise mysterious unity of identity and difference makes sense. Second, a phenomenon is also the unity of opposites, inasmuch as the potential features of a phenomenon are the (contradictory) potential opposite of its realised aspects. Third, a phenomenon is the unity of essence and appearance: its potential aspect is its essence, or that which manifests itself in a realised form, while its realised aspect is its (temporary and contingent) appearance, or the form taken by the possibilities inherent in its potential. The picture of social reality that emerges from dialectical thinking is a temporal flow of realised contradictory phenomena, either determinant or determined, and continuously emerging from and returning to a potential state. To explain change, we need dialectical logic. Formal logic pertains only to static reality.20
5 Dialectical Processes
The question now is whether dialectics pertains only to the social world or whether there is also a dialectics of nature. The answer requires that we distinguish between two types of dialectical process. In the pre-determined dialectical process, the realised form of a potential is already known, because it is the same as the former realised knowledge before it became a formless potential. In this case, the potential can take only that definite form, i.e. the outcome is pre-determined and known. We know the outcome beforehand, either with certainty (for instance, we know that water turns into ice) or with a certain probability (for instance, we know the probabilities of a coin toss).21 This also applies to mental processes, such as the repetitive application of an already known formula.
In the open-ended dialectical process, the realised outcome emerges from a formless potential as something new because it is different from the potential. An example is the solution to a problem, which is realised as new knowledge. To be new, knowledge must emerge from a formless state; it must become realised as something that did not exist previously in a definite, realised form. This is to say that the potential must be formless. How, then, can definite forms of knowledge emerge from formlessness, and how can the realised emerge from the potential? The answer is through a specific form of knowledge – that is, intuition.
Intuitive knowledge plays a fundamental role in the mental labour process, yet it is disregarded by most epistemologists – Marxist or otherwise. As Einstein famously said, ‘[t]he intuitive mind is a sacred gift and the rational mind a faithful servant’.22 Only intuition can provide ‘the creative generation of hypotheses, novel compositions of concepts, and novel discovery of reasoning rules’.23 Intuition taps from the reservoir of formless potentials – but why should this reservoir be formless? And why is this important for the explanation of the emergence, or realisation, of new knowledge?
Knowledge that enters the mental labour process is principally knowledge that has been absorbed by human labour power and subsequently become part of it. Different from knowledge that is contained within the objective means and objects of mental production (see Section 2, above), where knowledge is fixed in well-defined forms, knowledge contained within labour power is not an accumulation of forms of knowledge in their previously realised shape. The latter would imply a process of selection rather than one of the generation of new knowledge. Rather, when a realised form of knowledge is absorbed into labour power, it interacts with the previously accumulated forms of knowledge; it changes those forms while at the same time being changed by them. Through this interaction, the previously existing forms of knowledge lose their specificity and acquire the potentiality of something different. The knowledge accumulated in labour power is in a constant state of mutability. It is from this reservoir that realised forms of knowledge emerge. Recent research in AI supports the thesis of potential as formless. As Stephen Ornes remarks, the process through which the brain organises and accesses spatial information involves ‘recalling an entire network of memories and stored spatial data from tens of billions of neurons, each connected to thousands of others’.24
The above is needed to explain a fundamental feature of Marx’s epistemology: the social content of knowledge. Human consciousness implies that humans perceive, even if only at the intuitive level, the continuously changing myriads of social relations and processes that constitute society. Through the continuous influx of realised knowledge as potential knowledge into labour power, potential knowledge acquires its internally contradictory social content, which then emerges as realised knowledge with a new contradictory social content. The notion of socially neutral knowledge – without social determination and thus without social content – has a definite social content itself, which is that of denying social contradictions and, in the end, class contradiction in the production of knowledge. It follows that a mental labour process, usually referred to as thinking, creates knowledge that is new from both a cognitive and a social perspective. This is the first reason why computers cannot think, or why they cannot think like humans.
There is another reason why computers cannot think as humans do. Suppose, for the sake of reasoning, that the objections above could be overcome and that computers could indeed think and generate new knowledge. The point remains that thinking is not necessarily a creative activity. Chiselling a sculpture is a creative process that results in the objective realisation of a new idea (knowledge). Chiselling the copy of a sculpture is the outcome of a new mental labour process. The copy is obviously new, because it did not exist before it was chiselled into existence. However, the knowledge contained in the copy is not new; it is a repetition of an already existing knowledge.25 As a repetition, it was pre-determined. Only humans can think creatively and produce new knowledge, because only humans have a reservoir of formless potentials from which they can draw.
Mathematical formulae are oft-mentioned examples, usually seen as empty shells that have been created once and for all. Supposedly, what changes is their use. This forms the thesis of the social neutrality of knowledge. Consider 2 + 2 = 4, which is almost universally accepted as an example of a class-neutral form of knowledge, for it seemingly has no inherent class content. However, to begin with, 2 + 2 is not always equal to 4. In a system that progresses from 1 to 24, the formula holds. But in a system that progresses only from 1 to 2, 2 + 2 = 2. That aside, let us choose a frame of reference within which 2 + 2 is always equal to 4 and assume that 2 + 2 has been proved (or assumed) to be equal to 4. After the initial proof, that formula can be reused in the assumption of its validity.
It might therefore seem that its use does not require a mental labour process. This is not so, for its use each time is at the same time a new mental labour process. We re-create the formula (even if we do not create it ex novo) each time we use it, because new labour power and thus a new mental labour process are set in motion to arrive at the same, pre-determined outcome. However, even if the form is the same, its social content can change. The elements of knowledge that are absorbed into labour power change continuously, as does the social content of its formless potentials. Therefore, the computation of the formula is a pre-determined, dialectical mental labour process, in which the reproduction of the same formula (form) hides the production of its possibly changing social content. This explains the possibility for that formula to fit into, and be used within, a different social, class content. The social, class use of knowledge is determined not by its class neutrality but the social, class nature of its production. This is the contradictory nature of the elements (inputs) of its knowledge, their becoming formless, contradictory potentials of new knowledge and, finally, their emergence as new, realised knowledge with a specific social, class – and thus contradictory – content. Knowledge, whether new or reproduced, is never socially neutral.
The above is vital for a theory of transition, or for posing the question of whether and to what extent the means of production that exist under capitalism can be used during the transition to a socialist system.26 Such means of production incorporate knowledge that is informed by the capitalist production relation. A conveyor belt presents a typical case: its mechanised and repetitive movement denies workers their creativity and forces them to become appendices of machines. A new social system will require different means of production with a different social, class content. Here, there is a distinction from the example of a mathematical formula, in that the formula’s form does not change in a different social context, though its content can. The means of production, however, can only change their social, class content by changing their form.27
6 ‘Machine Thinking’, ‘Artificial Intelligence’ and ChatGPT
The question that new technologies force upon Marxist epistemology is whether machines – specifically computers, their software and AI – can think. A number of responses have been offered. Some hold that computers think because they can be made to learn. For others, computers think but cannot generate a new thought. There are some who argue that computers simply imitate human thought. There are others who simply assert that computers have no emotions. And finally, there are those who refute the possibility that computers can think because only human cognition is affected by social relations, processes and interests. There are valid points across each position. However, none gets to the heart of the matter, which is that even if computers could reason, they could not reason as humans do, i.e. dialectically.
There are three reasons why they cannot. The first is that computers lack a formless but socially determined potential. Instead, they hold a reservoir of knowledge that is constituted by already realised forms of knowledge. Computers can select and manipulate existing forms of knowledge, such as images, but they lack the human creativity that undergirds the capacity to think the previously unknown. It follows that computer software reduces qualitative differences to quantitative ones, purely in terms of what can be quantified and measured.
To unpack the second reason, we should first consider classical computers, whose transformations of knowledge are based on formal logic, on Boolean logic or on algebra, which denies the possibility that any one statement can be both true and false at the same time (see above). As formal logic excludes the possibility of contradiction, computers do likewise. Thus, if a contradiction could be perceived by a computer, it would be interpreted as a logical mistake. The same theory applies to quantum computers (see below). In brief, formal logic excludes open-ended processes; this is because it cannot draw on a formless store of knowledge which is internally contradictory due to the contradictory nature of the elements of knowledge sedimented into it. Different to formal logic, dialectical logic is based on contradictions, including the contradiction between the potential and the realised aspects of knowledge. This is the source of the contradictions between the realised aspects of reality, including elements of knowledge.
The above also holds for AI. Like computers, AI functions on the basis of formal logic. Take ChatGPT, for example. When asked if A = A and also A ≠ A at the same time, ChatGPT answers negatively. Because it uses formal logic to reason, AI cannot access potential knowledge and mine for new, contradictory knowledge. Coincidingly, it cannot conceive of contradictions, because it cannot conceive of inherently contradictory potential. Such contradictions are the humus of creative thinking that is required to unearth the yet-unrealised. AI can only select, recombine and duplicate realised forms of knowledge, but it can do so at a much greater speed than humans. In tasks such as image recognition, reading comprehension and game playing, it can perform much better than humans. It is a formidable means of mental production, but the fact remains that it cannot generate qualitatively new, in the sense of not yet realised, knowledge.
Consider facial recognition for another example, a technique that can compare the photograph of an individual with a database of images to find a match. The database consists of a number of other, known faces. Finding a match entails the selection of an already realised (that is, already known) face. This process involves no generation of new knowledge (via new faces). The facial recognition process of AI can find a match in much less time than a human can, and its utilisation can therefore make human labour more productive. But selection is not creation. Selection is a pre-determined mental process; creation (which involves a change of perspective) is an open-ended mental process.
ChatGPT would seem to emulate the creative writing of humans. In actuality, it does not. Rather, it receives knowledge from large amounts of text data (comprised of the objects of mental production), which it divides into smaller pieces, phrases, words and syllables – the so-called tokens of its generative system. When ChatGPT ‘writes’, it does not produce a sequence of tokens according to the logic of an argument (as humans do). Instead, it chooses the most likely token based on the frequency in which it appears in association with other tokens in the already available data. The written outcome is a chain of tokens which are assembled on the basis of the statistically most probable combination. If it is based on the most frequently used combinations, chances are that the generated text carries some logical argument (though this is not always the case). Nevertheless, the outcome is formed through a selection and combination of elements from already existing knowledge; it remains that this process is not the creation of new knowledge. As Chomsky et al. put it: ‘AI takes huge amounts of data, searches for patterns in it and becomes increasingly proficient at generating statistically probable outputs – such as seemingly human-like language and thought. … [ChatGPT] merely summarizes the standard arguments in the literature’.28
Morozov provides another relevant example:
[Consider] Marcel Duchamp’s 1917 work of art Fountain. Before Duchamp’s piece, a urinal was just a urinal. But, with a change of perspective, Duchamp turned it into a work of art. When asked what Duchamp’s bottle rack, the snow shovel and the urinal had in common, ChatGPT correctly answered that they are all everyday objects that Duchamp turned into art. But when asked which of today’s objects Duchamp could turn into art, it suggested smartphones, electronic scooters and face masks. There is no hint of any genuine ‘intelligence’ here. It’s a well-run but predictable statistical machine.29
In sum, the ‘new’ is not a reworking of the past. ChatGPT is a mental means of mental production. Like all knowledge under capitalism, it has a value: the variable and constant capital that has gone into its production plus the generated surplus value. It participates in the tendential formation of society’s average rate of profit. Its amortisation is determined by the emergence of alternative and competing means of mental production, which eventually cause it to become obsolete. For instance, ChatGPT3 is an obsolete version of ChatGPT4.
GPT and AI in general, cannot reason, as humans do, on the basis of causal connections. Because of this – or because they function on the basis of the probabilities of occurrences –, and in the absence of continuous human intervention, AI inevitably makes mistakes that humans do not and cannot make (unless they do so intentionally). This occurs in the progressive loss of information (or knowledge) each time an AI model is fed more AI-generated data in the absence of human-produced data. The reason for this is the necessarily restrictive programming of models that are based on the probability of occurrence; these models focus on and select data with the highest probability, thus overlooking much other information. It follows that each time less-comprehensive datasets are fed into the model in a recursive way, the model loses more information about less probable occurrences and thus generates a distorted view of reality. Such is the process of so-called model collapse.30 For example, we can think of the hypothetical scenario
wherein a machine learning model is trained on a dataset with pictures of 100 cats – 10 of them with blue fur and 90 with yellow. The model learns that yellow cats are more prevalent, but also represents blue cats as more yellowish than they really are, returning some green-cat results when asked to produce new data. Over time, the original trait of blue fur erodes through successive training cycles, turning from blue to greenish, and ultimately yellow.31
The changed data distributions arise from the model itself as a result of previous training (or the input of data).
It follows that, to avoid model collapse, access to genuine, human-generated content is essential. As Shumailov et al. argue, there are two methods for avoiding data degeneration. The first is by retaining a copy of the original, human-produced dataset. This would mean that the model could be periodically retrained, starting from scratch with the original data. The second method is to introduce new human-generated datasets into the model rather than reintroducing the old, original data.32 In both cases, human intervention is essential. Without it, the AI model, training only on machine-generated data, collapses in an outcome shown by Shumailov et al. to be inevitable.33
In short, AI differs fundamentally from the human generation of knowledge. This is because (a) the interconnections drawn by AI are based on probabilities of occurrence rather than causal reasoning; (b) the knowledge AI produces is based on formal rather than dialectical logic; (c) AI incorporates no sight of the social determinations of knowledge production; and because (d) left to their own devices, AI models deteriorate and inevitably collapse. Therefore, common fears that computers and AI might overpower and subdue humans are undergirded by childish fantasy and nothing more.
Two points of clarification must follow the above arguments. First, according to Marx, besides being unique concrete individuals, humans are also abstract individuals, or carriers of social relations. As an abstract individual, ‘the human’ is a general designation which obliterates differences between individuals, social groups and classes, including their different interests and world views. Even if machines (such as computers) could think, they could not think like class-determined humans with different, class-determined conceptions of what is true and what is false, or of what is right or wrong. The belief that computers are capable of emulating human thought is not only wrong, it is also a pro-capitalist ideology because of its blindness to the class content of the knowledge that is stored in human labour power, which is thus revealing of its incapacity to grasp the contradictions that are inherent to the generation of knowledge.
Second, formal logic pertains only to the realm of the realised. It cannot be used to think of potentiality or of the previously unknown.34 What escapes this logic’s compass is that realised phenomena are continuously subjected to dialectical change: their status continuously shifts from being the determinant to the determined, and vice-versa; from tendential to counter-tendential, and vice-versa; and from potential to realised, and vice-versa. These changes are beyond formal logic’s domain and thus beyond the perspectives of those who live in blissful ignorance of dialectics. Both types of logic, formal and dialectical, are needed for a full analysis of reality, though this rests on the condition that one knows that formal logic only applies when the potentialities inherent in reality are divorced from consideration.
7 ‘Quantum Thinking’
The section above has dealt with knowledge of the macro world. This section focuses on quantum theory, whose field of research is the sub-atomic world. Due to the increasing importance of its applications (for example, in quantum computers), Marxist epistemology cannot ignore this area of thought any longer.
Quantum theory’s mathematical formalism is beyond the reach and scope of this paper’s work. Rather, focus will remain on the philosophical and ideological foundations of the theory’s mainstream interpretation, known as the ‘Copenhagen interpretation’ (hereafter, CI).35 What follows is by no means an exhaustive retelling of this CI. Rather, it hones in on one of the CI’s aspects – namely, superposition. In Schrödinger’s words, superposition is not simply one aspect but the characteristic trait of quantum mechanics’.36 This limited focus is sufficient to show the CI’s irrationalism and class content. Irrationalism is here defined as the belief in something that cannot be proved to exist by definition.
Let us consider a game of heads or tails, which presents a pre-determined dialectical process. This is a macro process. As such, it does not apply to the sub-atomic world, although it is used in popularising literature as an intuitive illustration of superposition. For the CI, as long as the coin spins, heads and tails do not exist as such but in a suspended or superposed position – that is, they exist as neither heads nor tails. What they are in this suspended state, we do not know, and the CI does not (because it cannot) enlighten us. In actuality, however, while it is rotating, the coin is potentially both heads and tails; when the movement stops, it is either heads or tails. It is, then, never neither heads nor tails, as the CI holds. So, something that is, for the CI, a mysterious state beyond reality is, in actuality, a potential state that is firmly rooted in reality.
The usual assumption is that the coin falls on either of the two sides. However, it could fall on the edge that divides heads from tails, not horizontally but vertically. In this case, in throwing a coin, there are not two but three potential outcomes. The third is usually disregarded, as the probability of its occurrence is very small. Nevertheless, not only can it occur, its potentiality is also very important theoretically.
The coin’s complete revolution consists of two cycles. The first goes from a horizontal position (say, heads) to a vertical position; the second goes from the vertical position to the other horizontal position (say, in this case, tails). Therefore, the vertical position simultaneously marks the end of the first cycle and the beginning of the second cycle. In the vertical position, the two states are not superposed. They do not exit this dimension to migrate instantaneously to another unknown and unknowable one, then to reappear instantaneously in this one. Rather, in the vertical position, heads and tails are frozen as two potential states. This is a real, objective, temporal and pre-determined process, a process through which a realised state is becoming a qualitatively different realised state. This has nothing to do with the irrationality and subjectivism of the CI.37
It is possible that Marx would have put forward the same criticism of superposition. In his mathematical manuscripts, he deals with the infinitesimally small, which is an infinite approximation to zero, and argues that something which, ‘as a realized entity is neither a number nor zero, should be rejected as “metaphysical”, as a “chimera”’.38 For Marx, there is no universe populated by superposed entities, which are neither numbers nor zero, or entities that become realised as a number or zero when they collapse into this world. Marx would thus would have called this notion metaphysical and ‘chimerical’.
Let us consider now an example from the subatomic world: the well-known two-slit experiment.39 Its purpose is to determine whether electrons are waves or particles. In the experiment, an electron is shot through a barrier with two slits. If only one slit is open, its hits are recorded on a photographic screen behind the barrier. If both slits are open, one would expect that each particle would pass through only one of the two and then hit the screen directly behind, i.e. we would expect just two points on the screen, one for each slit. However, if an electron is shot repeatedly, a scattered form appears which is called a wave. So, the question is: is an electron a particle or a wave? If a detector is put on each slit, we can observe that the pattern is the same as it would have been if only one slit were open, but doubled – that is, we can find one such pattern for each slit.
How does the CI interpret this result? Before the measurement, the electrons are in a superposition state, in a wave pattern, which means that they can pass through each slit. After measurement, their superposition collapses and each electron passes through one of the slits.
The hypothesis that two states exist in superposition and that one is realised at the moment of measurement is unprovable by definition. The argument that it is necessary to account for reality also falters: other hypotheses can do the same – for example, a creationist act of God. As Ball remarks, ‘This collapse was first proposed by John von Neumann in the 1930s. … [It] was a sleight of hand with no real justification from the theory itself: a mathematical convenience to reconcile the theory with what we actually see’.40
Nevertheless, some authors attempt to square the circle. For example, Wendt submits that: ‘before measurement, quantum systems are potentialities, yet afterward we get actualities’.41 However, we know neither what these potentialities are nor how they become actualised through measurement and observation. Moreover, as de Ronde and Massri argue, superposed states ‘might or might not become actualized in a future instant of time. … [Therefore, they] cannot be considered as elements of physical reality’.42 If superposed states are not physical reality, what are they?
Ball comes closer to a more rational view: ‘A quantum particle in a superposition, contrary to common belief, is not really in two (or more) states at once. Rather, a superposition means that there is more than one possible outcome of a measurement’.43 One could advance some possible rational reasons for this. For example, it could be suggested that adding detectors changes the measuring device and thus the outcome of the measurement. Why, then, have physicists chosen superposition and, more generally, irrationality for rationality? An attempt to answer this question is provided below, in Section 9.
8 Probability versus Potentiality
For Werner Heisenberg, a founder of quantum mechanics and an adherent to the CI, the concept of the probability wave ‘was a quantitative version of the old concept of “potential” in Aristotelian philosophy’.44 Furthermore, Heisenberg posited that ‘[t]he probability wave function … meant a tendency for something’.45 That said, potentialities are not an abstract mathematical space of probabilities but part of this reality. Otherwise, how could an abstract mathematical space of probabilities become something concrete and objective? Again, the CI is not equipped to offer an explanation for such a question.
The relation between probability and potentiality should be framed within the dialectical approach. In the pre-determined process of tossing a coin, heads and tails exist contemporaneously – not in a superposed state but as two separate, potential states, or as moving reality. The emergence of either side depends on a number of factors. For instance, if the initial state is heads and the revolving process is stopped shortly thereafter, chances are that heads will be the realised state. As such, probability measures the likelihood of the realisation of a potential, or of something real. However, this principle only holds for pre-determined processes; it does not apply for open-ended processes or thus for the generation of knowledge. For that reason, no relation can be said to exist between potentialities and probabilities, because probabilities cannot measure something that has no form.
Contrary to superposition, we know what potentialities are even if they are formless, like a painting one is attempting to put on the canvas. Potentialities are sedimentations of material processes (including processes of mental production) immersed in time and in changing social and class relations and processes. In contrast, we do not know – and there is no way in which we might come to know – what superpositions are before they collapse, where they come from or why observation makes one collapse and not another. Most importantly, we do not know the process through which superpositions emerge in this dimension. Of course, one can renounce the understanding of what is measured and retain the description and measurement of the unknown and unknowable. To do so would align with Bohr’s view that ‘[t]here is no quantum world. There is only an abstract quantum physical description’.46 This is also the view that is promoted by the CI. Without attempting to answer the questions posed above, it simply nods to the impossibility of answering them at all.
For the CI, superposed entities exist in an a-temporal dimension. If they are a-temporal, their collapse is instantaneous. But if collapse is instantaneous, there must be a part of the universe – a sub-atomic space – where time does not exist. However, given that a lack of time implies a lack of movement, we can ask: if the universe is expanding, and thus moving, how could a part of it be immobile? A problem is thus unveiled for the astrophysics’ observation that the universe is expanding. Furthermore, why is there no time in the sub-atomic world? How can timelessness relate to time-full reality?
If the sub-atomic world is in a state of superposition, many parallel universes would have to exist. Indeed, this scenario substantiates the fashionable Many-Worlds interpretation, which the most fanciful adherents of the CI also extend to the macro world.47 Still, because the other worlds (and how many could there be?) cannot be known before one of them collapses into this one, it is impossible to show that other worlds exist or, if they do, what they are. It could thus be argued that, because it cannot be observed, dark matter cannot be proven to exist either. However, it can be theorised, and we can observe the ‘influences it exerts on the luminous matter’.48 Thus, it is possible to hypothesise what dark matter is and to prove one’s own hypothesis. On the other hand, the existence of many worlds cannot be proven by definition.
Timelessness and thus irrationality are the basic features of superposition and thus of the CI. Its problems disappear as soon as superpositions are replaced by potentialities. Given that potentialities are a formless reservoir of many possible realisations, different realisations can emerge from the same formless potential – not only sequentially but also at the same time. Here, time is not cancelled. Rather, the contemporaneous realisation of two outcomes from the same reservoir of potentialities is the outcome of the same time-full process of transformation.
The above critique applies also to the CI of quantum computers. If classical computers cannot generate new knowledge, is it not possible that quantum computers could, given their much greater computational power? To argue as much would depend on the assumption that new knowledge arises from a vastly greater computational speed, but this supposition has been disproven.
One does not need to know how quantum computers work, for there is no question that they work. It is the CI that is faulty. As Daniel puts it: ‘the usual formalism of quantum mechanics can be recovered, but within a consistently realistic interpretation’.49 Indeed, the realistic interpretation submitted here substitutes potentialities for superpositions; realism and materialism for unrealism and idealism; and the all-pervasiveness of time for timelessness. In short, dialectical determination substitutes both rigid causality and statistical correlations.
9 Irrational Thinking
Finally, it is worthwhile to ponder upon the reasons why some exceptionally talented physicists could (and still) resort to irrational and mystical interpretations of their ground-breaking achievements.
Forman provides a basic starting point, having found ‘overwhelming evidence that in the years after the end of the First World War but before the development of an acausal quantum mechanics’ – and under the influence of ‘the intellectual milieu in which the German physicists were working and quantum mechanics was developed’ – ‘large numbers of German physicists, for reasons only incidentally related to developments in their own discipline, distanced themselves from, or explicitly repudiated, causality in physics’.50 Why did they do so?
Physicists, like other members of society, interact with their intellectual milieu. This is why most of them could switch from a causal to an acausal, probabilistic way of thinking.51 What, then, was that intellectual environment that provided the ground for such a shift? It was one suffused with a sense of crisis in science, and an explicitly anti-rationalistic attitude towards it, as well as a generalised acceptance of irrationalism and mysticism. Thus, ‘the same scientist who in one context offers resistance to the antiscientific currents of his milieu, in another context can be found flirting with propositions intimately associated with those same currents’.52 For Forman, the reason behind many physicists crossing over to a-causality was psychological. He comments that ‘[t]he German physicists’ predisposition towards acausal laws of nature … arose as a form of accommodation to their intellectual environment’.53 For scientists who were ‘anxious to play up to that audience’,54 clinging to causality meant losing prestige and risking marginalisation.
The conversion to a-causality was greatly facilitated by the erroneous identification of causality with absolute determinism, a view that laws hold without exception. For dialectical determination, on the contrary, there are not exceptions but countertendencies. However, such an understanding was beyond the reach of the Weimar physicists – this being why, to reject rigid determinism, they saw no option but to embrace probabilism.55
Forman’s work has the merit of stressing the influence of the Weimar intellectual milieu on German physicists. It also underscores the contradictory class content of that milieu. It is this contradictory content that explains why some physicists drifted from a-dialectical causality while others did not. Such divergences in turn clarify both the role of the individual in the formation of the class-determined intellectual milieu as well as how the individual is shaped by it, as part their social environment.56
Hassani steers further the research towards rationalism and causality: ‘the quantum physics–mysticism association goes back to the founders of quantum physics themselves, and the convergence of three factors facilitated that association: the infiltration of Eastern thought in Western philosophy, the rise of mysticism in the West, and the unique character of quantum physics’.57 But how did irrationalism penetrate Western philosophy? ‘One of the unintended consequences of [classical] physics was the eventual decline in the traditional Western religions. … [This decline] created a moral vacuum by the end of the nineteenth century that could be filled only by a belief system [based on irrationalism]’.58 And as Hassani further specifies, ‘the founders of quantum physics – Niels Bohr, Werner Heisenberg, Wolfgang Pauli, and Erwin Schrödinger – all developed a strong affinity for Eastern theosophy and, regrettably, tied their science to that mystical viewpoint’.59 The reliance of the CI on philosophical thought that was itself tainted by irrationalism was the channel through which irrationalism penetrated and shaped the CI.
Hassani mentions the influence of Schopenhauer on both Pauli and Schrödinger. In The Destruction of Reason, Lukács depicts Schopenhauer’s ‘religious atheism’ in terms that are strikingly similar to the CI. He states that ‘for those no longer able to believe in the dogmatic religions’, Schopenhauer’s view matched ‘scientific requirements on the one hand and “metaphysical” needs on the other … broadly accommodating the lingering emotional attachment to religious or semi-religious prejudices’.60 Furthermore, for Schopenhauer, ‘the mode of comprehension … is incapable in principle of telling us anything about the phenomena’s essence’.61 Indeed, this incapability is also a feature of superposition, which is an example of modern irrationalism, or the idea that scientific laws are abstractions that explain the behaviour of phenomena but do not reveal their essence.62
Hassani’s view that the irrationalism of the CI was determined by the decline of Western religions provides a key insight. This, however, does not get to the bottom of the matter that is the social and class determination of the irrationalism inherent in the CI. For this, Lukács provides the other crucial insight, stating that ‘the various stages of irrationalism came about as reactionary answers to problems to do with the class struggle’.63
The origin of quantum theory did coincide with some of the most acute moments of class strife in the history of Western societies. To be specific, it emerged during the upsurge of worker militancy around 1919–20 as a result of the tidal uprisings and revolutions around the world, which were sparked by the victory of the October Revolution and fuelled by the aftermath of World War I.64 It also emerged in the context of capital’s reaction to those revolutions: a policy of containment based not only on brutal, violent and homicidal repression but also on the spread of irrational conceptions throughout society (including large sections of labour), such as fascism’s notion of white supremacy or of the genetically-determined egoistic nature of humans.
Despite its many different forms, irrationalism in modern capitalist societies has a precise class content, which is that of diverting labour’s attention from the nature and reality of capitalism itself – and thus also from the rationality of labour’s struggle against it. Seen from this angle, perhaps Lukács’s book should have been called the Destruction of Labour’s Rationality instead of The Destruction of Reason. From labour’s perspective, the antithesis is between capital’s irrationalism and labour’s rationalism. Irrationalism is a weapon, besides brute force, that has been deployed to win the minds of labourers. It penetrated labour’s collective consciousness and became most poisonous when it was most needed by its wielders – which is to say, upon the birth of fascism in Italy in 1921 and later upon the rise of National Socialism in Germany. It is because of its class content that irrationalism persists today as the decisive principal stream of reactionary philosophy.65 As for the CI, its irrationalism is its belief in superpositions, in the existence of a dimension that cannot be known, where time and change do not exist. This irrationality has shaped, keeps shaping and thus reproduces the CI because of its social content, irrespective of the failures of individual physicists to recognise the fact. In such a way, the CI plays into the hands of capitalism, even if inadvertently. It is a pro-capitalist ideology, even if it does not directly address itself to social and class strife. The practical success of quantum theory, however, lends credibility to the CI’s irrational interpretations and reinforces them.
To say that capitalism generates irrational thoughts is to present a convenient short-cut that should be unpacked. Capitalism is not an individual but a system of social relations among individuals. How, then, could some of the founders of quantum physics respond to the needs of capital and develop an irrational interpretation of their own work? More generally, how can individuals become ideological representatives of social classes, irrespective of their own position in the social structure, personal convictions and beliefs?
This is one of the pivotal questions of Marxist epistemology, which can only be touched upon briefly.66 Taking a wrong approach, Erik Olin Wright first defined class structure not in terms of production relations but in Weberian terms – namely, in terms of personal occupation, income, status and so on. He then conceptualised class consciousness (as pro-labour or pro-capital) on the basis of eight questions and posed them to a representative sample of individuals. Wright finally applied statistical procedures to his survey to determine whether positions in the class structure map onto a pro-capital or pro-labour consciousness.67 Can a Marxist ideology and epistemology be developed on these grounds?
Wright’s questions were detached from their social context and thus could not reveal class consciousness. For instance, one question contained the statement that ‘corporations benefit owners at the expense of workers and consumers’. According to his statistical rubric, a respondent was deemed to reveal a pro-labour consciousness if they agreed with this statement. Nevertheless, a fascist or populist respondent may have also answered positively without being of a pro-labour disposition. Moreover, without assessing the prominent roles of the relations of production and of class struggle in the formation of individual consciousness, it is impossible to account for how the world perceptions and political inclinations of those same individuals could have been radically altered.
As discussed in Section 6, Marx holds that humans are always and at the same time both unique individuals (or concrete and thus specific individuals) and – by virtue of living in society – carriers of contradictory social relations (or abstract individuals). Despite their uniqueness, human individuals share common features because they are carriers of the same social relations. They relate to each other on the basis of those features and thus possibly contradictorily. Common features are the factors that aggregate unique individuals into social groups and classes, under specific historical circumstances and with different interests. Therefore, contrary to Wright’s judgements, it does not matter whether an individual labourer’s worldview is consonant with the view of capital or of labour as defined by the terms of eight questions. At the individual level, it is a matter of chance as to which ideology is adhered to. However, chance at the individual level is necessity at the social, class level. As long as economically determined classes exist, so too will unique individuals who embody a certain class ideology – whether the individual’s consciousness maps onto their economic position or not.68 This claim applies to physicists and their relations in society at large. Some, as unique individuals, absorb irrational ideologies, which are then transformed into irrational explanations of scientific findings.
To summarise, Forman, Lukács and Hassani provide the co-ordinates with which the CI’s irrationalism can be located. Forman highlights the role of the intellectual milieu in shaping the CI, while Hassani details irrationalism’s penetration of the CI at around the same time. The Zeitgeist of that time was strongly permeated with irrationalism, which pervaded the consciousness of philosophers and, through this channel, the irrational perceptions of the work of physicists. Lukács explains the rise of irrationalism in terms of its class content – not only as an offspring of capitalism, but as a most-important one.69 When the CI was born, irrationalism was a powerful instrument to defend and foster the interests of the capitalist class, whether the physicists were aware of this or not. Due to its class content, the ideological function of the CI lives on today.
10 Conclusion
This article has focused on a largely unexplored and yet crucial area of Marxist epistemology: the extension of the objective (wrongly called the material) labour process, as in Marx, to the mental labour process, i.e. the production of knowledge. This category rests on two pillars. The first concerns the materiality of knowledge: if knowledge is material, the basic distinction is not between material and mental but between objective and mental aspects of reality. Both are material. The second pillar is that of dialectics, which is here sought in Marx rather than in Hegelian philosophy. Dialectics emphasises the double dimension of reality: the realised and the potential. This view accounts for real processes from a dialectical perspective, specifically the dialectic of material transformation (both objective and mental) between the determinant and the determined; the tendential and the counter-tendential; the potential and the realised. Matter is thus seen in a state of continuous movement. But movement also implies time. Therefore, all theories – in both the physical and social sciences – which neglect considerations of time are at best useless.
The outcome of a mental labour process can be either pre-determined or open-ended. Only the latter can explain new knowledge, or the previously unknown. Predetermined labour processes can only reproduce previous knowledge, if at a much higher speed, as in AI. This, however, is not creative thinking. Computers can only engage in predetermined labour processes, i.e. using formal logic. This is one of the reasons why computers cannot think in the manner of humans.
Finally, the analysis of the mental labour process in its double nature, as a material as well as a socially determined process, dispels the myths of the social neutrality of knowledge, including that of the knowledge frozen in the objective means of (mental) production. This is of vital importance for a theory of transition to socialism.
References
Ananthaswamy, Anil 2023, ‘Double Trouble: Two Slits, Many Questions’, Nature, 618: 454–456.
Ball, Philip 2021, ‘How Big Can the Quantum World Be? Physicists Probe the Limits’, Quanta Magazine, available at: <https://www.quantamagazine.org/how-big-can-the-quantum-world-be-physicists-probe-the-limits-20210818/>.
Bohr, Aage, Ben R. Mottelson and Ole Ulfbeck 2004, ‘Quantum World Is Only Smoke and Mirrors’, Physics Today, 57, 10: 15–16.
Blunden, Andy 1984, ‘Dialectics and Mathematics’, Labour Review, available at: <https://ethicalpolitics.org/ablunden/pdfs/Dialectics_Mathematics.pdf>.
Brynjolfsson, Erik 2022, ‘The Turing Trap: The Promise & Peril of Human-Like Artificial Intelligence’, Daedalus, 151, 2: 272–287.
Burke, Amy, Abigail Okrent and Katherine Hale 2022, ‘The State of US Science and Engineering 2022’, Science and Engineering Indicators, available at: <https://ncses.nsf.gov/pubs/nsb20221>.
Carchedi, Guglielmo 1975, ‘On the Economic Identification of the New Middle Class’, Economy and Society, 4, 1: 1–86.
Carchedi, Guglielmo 1984, ‘The Logic of Prices as Values’, Economy and Society, 13, 4: 431–455.
Carchedi, Guglielmo 1989, ‘Class and Class Analysis’, in Approaches to Class Analysis, edited by Erik Olin Wright, Cambridge: Cambridge University Press.
Carchedi, Guglielmo 2008, ‘Dialectics and Temporality in Marx’s Mathematical Manuscripts’, Science & Society, 72, 4: 415–426.
Carchedi, Guglielmo 2010, Behind the Crisis: Marx’s Dialectic of Value and Knowledge, Historical Materialism Book Series, Leiden: Brill.
Carchedi, Guglielmo 2012, ‘Mathematics and Dialectics in Marx: A Reply’, Science & Society, 76, 4: 546–549.
Carchedi, Guglielmo 2014, ‘Old Wine, New Bottles and the Internet’, Work Organisation, Labour and Globalisation, 8, 1: 69–87.
Carchedi, Guglielmo 2022, ‘The Ontology and Social Dimension of Knowledge: The Internet Quanta Time’, International Critical Thought, 12, 4: 597–626.
Carroll, Sean M. 2019, Something Deeply Hidden: Quantum Worlds and the Emergence of Spacetime, London: Oneworld Publications.
Charig, Nigel and Nicole Kareta 2020, ‘Quantum Computers: How Do They Work, and What Might We Expect Of Them?’, Power & Beyond, available at: <https://www.power-and-beyond.com/quantum-computers-how-do-they-work-and-what-might-we-expect-of-them-a-c6371d5d640900f569cfcf719816135a/>.
Choi, Yejin 2022, ‘The Curious Case of Commonsense Intelligence’, Daedalus, 151, 2: 139–155.
Chomsky, Noam, Ian Robert and Jeffrey Watumull 2023, ‘Noam Chomsky: The False Promise of ChatGPT’, New York Times, 8 March, available at: <https://www.nytimes.com/2023/03/08/opinion/noam-chomsky-chatgpt-ai.html>.
Cross, Andrew 1991, ‘The Crisis in Physics: Dialectical Materialism and Quantum Theory’, Social Studies of Science, 21, 4: 735–759.
de Ronde, Christian and César Massri 2019, ‘Against the Tyranny of “Pure States” in Quantum Theory’, Foundations of Science, 27: 27–41.
de Ronde, C., G. Domenech and H. Freytes, ‘Quantum Logic in Historical and Philosophical Perspective’, Internet Encyclopedia of Philosophy, available at: <https://iep.utm.edu/qu-logic/>.
Engels, Friedrich 1987 [1883], ‘Dialectics of Nature’, in Marx/Engels Collected Works, Volume 25, Moscow: Progress Publishers.
Forman, P. 1971, ‘Weimar Culture, Causality, and Quantum Theory, 1918–1927: Adaptation by German Physicists and Mathematicians to a Hostile Intellectual Environment’, Historical Studies in the Physical Sciences, 3: 1–115.
Frank, Adam and Ethan Siegel 2022, ‘Is the Multiverse Real? Two Astrophysicists Debate’, Big Think, available at: <https://bigthink.com/13-8/is-the-multiverse-real/>.
Franzen, Carl 2023, ‘The AI Feedback Loop: Researchers Warn of “Model Collapse” as AI Trains on AI-generated Content’, Venture Beat, available at: <https://venturebeat.com/author/carlfranzen/Venture Beat>.
Gerla, Giangiacomo 2017, ‘Vagueness and Formal Fuzzy Logic: Some Criticisms’, Logic and Logical Philosophy, 26: 431–460.
Ghirardi, Giancarlo and Angelo Bassi 2020, ‘Collapse Theories’, Stanford Encyclopedia of Philosophy, available at: <https://plato.stanford.edu/archives/sum2020/entries/qm-collapse/>.
Hassani, Sadri 2020, ‘Sources of Quantum Voodooism’, Skeptical Inquirer, 44, 6, available at: <https://skepticalinquirer.org/2020/11/sources-of-quantum-voodooism/>.
Henninger, Max 2007, ‘Doing the Math: Reflections on the Alleged Obsolescence of the Law of Value under Post-Fordism’, Ephemera, 7, 1: 158–177.
Howard, Don 2014, ‘Who Invented the “Copenhagen Interpretation”? A Study in Mythology’, Philosophy of Science, 71, 5: 669–682.
Joque, Justin 2022, Revolutionary Mathematics: Artificial Intelligence, Statistics and the Logic of Capitalism, London: Verso.
Kangal, Kaan 2017, ‘Carchedi’s Dialectics: A Critique’, Science and Society, 81, 3: 427–436.
Kliman, Andrew 2007, Reclaiming Marx’s Capital: A Refutation of the Myth of Inconsistency, Lanham, MD: Lexington Books.
Kojevnikov, Alexei 2013, ‘Probability, Marxism, and Quantum Ensembles’, Yearbook of the European Culture of Science, 6: 211–235.
Kosko, Bart and Satoru Isaka 1993, ‘Fuzzy Logic’, Scientific American, 269, 1: 76–81.
Kostakis, Vasilis 2010, ‘Identifying and Understanding the Problems of Wikipedia’s Peer Governance: The Case of Inclusionists versus Deletionists’, First Monday, 15, 3, available at: <https://firstmonday.org/ojs/index.php/fm/article/download/2613/2479>.
Lenin, Vladimir 1972 [1908], ‘Materialism and Empirio-criticism’, in Collected Works, Volume 14, Moscow: Progress Publishers.
Lobkowicz, Nikolaus 1961, ‘The Principle of Contradiction in Recent Soviet Philosophy’, Studies in Soviet Thought, 1, 1: 44–51.
Lukács, Georg 2021 [1952], The Destruction of Reason, translated by Peter Palmer, London: Verso.
Marx, Karl 1967, Capital, Volume I, New York: International Publishers.
Marx, Karl 1983, Mathematical Manuscripts of Karl Marx, New York: New Park Publications.
Morozov, Evgeny 2023, ‘The Problem with Artificial Intelligence? It’s Neither Artificial nor Intelligent’, The Guardian, 30 March, available at: <https://www.theguardian.com/commentisfree/2023/mar/30/artificial-intelligence-chatgpt-human-mind>.
Ornes, Stephen 2022, ‘How Transformers Seem to Mimic Parts of the Brain’, Quanta Magazine, available at: <https://www.quantamagazine.org/how-ai-transformers-mimic-parts-of-the-brain-20220912/>.
Sayers, Sean 1983, ‘Materialism, Realism and the Reflection Theory’, Radical Philosophy, I, 33: 16–26.
Selsame, Howard (ed.) 1949, Handbook of Philosophy, adapted from the Short Philosophic Dictionary by Mark Moisevich Rosenthal and Pavel Fyodorovich Yudin, New York: International Publishers.
Shumailov, Ilia, Zakhar Shumaylov, Nicolas Papernot, Yiren Zhao, Yarin Gal and Ross Anderson 2023, ‘The Curse Of Recursion: Training on Generated Data Makes Models Forget’, arXiv, available at: <https://arxiv.org/abs/2305.17493>.
Starosta, Guido 2012, ‘Cognitive Commodities and the Value Form’, Science & Society, 76, 3: 365–392.
Wendt, Alexander 2015, Quantum Mind and Social Science: Unifying Physical and Social Ontology, Cambridge: Cambridge University Press.
Wright, Erik Olin et al. 1989, The Debate on Classes, London: Verso.
This article is a substantive revision of Part One of Carchedi 2022.
We shall see shortly that knowledge is the outcome of mental and objective transformations in which the former is determinant. At this stage, focus remains on the materiality of mental transformations. If mental transformations are material, and if objective transformations are too, then knowledge cannot be anything but material.
Burke, Okrent and Hale 2022. R&D is the fundamental tool of international rivalry, especially between the US and China. According to Burke, Okrent and Hale, ‘[b]ased on R&D expenditures, a few countries perform most of the global R&D. In 2019, the United States (27% or $656 billion) and China (22% or $526 billion) performed about half of the global R&D’.
A complete critique of workerism is beyond the scope of this article. See Carchedi 2011; Henninger 2007; Starosta 2012. For some authors, especially of a workerist persuasion, ‘immaterial’ seems to carry the same meaning as ‘mental’ does in the present work. This is a superficial resemblance, as workerism rejects Marx’s labour theory of value while the present work retains it as a solid basis for Marxist epistemology.
Lenin 1972, p. 192.
Marx uses the term ‘reflection’ in a very concise paragraph to explain the essential difference between his epistemology and that of Hegel: ‘To Hegel … the real world is only the external, phenomenal form of “the Idea”. … With me, on the contrary, the ideal is nothing else than the material world reflected by the human mind, and translated into forms of thought’ (Marx 1967, p. 19; emphasis mine). Traditionally, commentators have focused on ‘reflection’ and ignored the transformation that is implied by ‘translation’.
Sayers 1983, p. 20.
Selsam (ed.) 1949, p. 56.
As quoted in <https://encyclopedia2.thefreedictionary.com/>.
A computer is an objective means of mental transformation, and the knowledge contained within it is a mental object of mental transformation.
See Carchedi 2011, Chapter 1.
The ontology of dialectics is the subject of the next section.
This point is particularly important within the context of the transformation debate in economic theory. See Carchedi 1984, Carchedi 2012, and further discussion below.
Brynjolfsson 2022, p. 277. See also Carchedi 2022.
The difference between knowledge and information is crucial. Information pertains only to the world of quantities and not to the qualitative, social aspects contained in the world of potentials.
For a more detailed treatment, see Carchedi 2011, pp. 39–44.
Engels 1987, p. 49.
Marx 1967. See Chapter 6.
We shall see that the potential is formless only for open-ended dialectical processes and not for pre-determined processes.
See Blunden 1984.
In Carchedi 2008 and Carchedi 2011, I argued that natural phenomena are not dialectical processes, because human consciousness and volition play no role in them. This was my basic critique of Engels’s dialectics of nature. In 2017, Kangal took issue with this stance. The difficulty is solved by introducing the difference between predetermined and open-ended processes. Also, the notion of dialectics as the science of universal interconnections did not seem to me to be very helpful. If everything is interconnected, and everything is thus dialectical, it becomes impossible to conceptualise the un-dialectical thinking of formal logic. However, the fundamental critique I was not able to make at that time is that universal interconnectedness does not penetrate the core of Marx’s analysis, i.e. potentiality and its relation to realisations. What characterises Marx’s dialectics is the realised–potential–realised double transformation. This applies both to the predetermined dialectical processes (which include Engels’s dialectics of nature and the repetition of already-known knowledge, but which do not include the generation of new knowledge) and to the open-ended dialectical processes (which include the generation of new knowledge, in the sense of the previously unknown).
Source: <https://quoteinvestigator.com/>.
Choi 2022, p. 144.
Ornes 2022.
This presupposes an exact replica.
See Carchedi 2014 and Carchedi 2022.
There are exceptions, such as those means of production that can fit into different sets of social relations and, thus, relations of production (for example, a hammer). Only class analysis can determine which means of production can be categorised as exceptional in such a way.
Chomsky, Robert and Watumull 2023.
Morozov 2023.
Franzen 2023.
Ibid.
Shumailov, Shumaylov, Papernot, Zhao, Gal and Anderson 2023.
Ibid.
There is a branch of logic – ‘fuzzy logic’ – that claims to reproduce human thinking better than classical binary logic. This is so claimed because the human brain can reason with vague assertions or claims that involve uncertainties or value judgements (Kosko and Isaka 1993). In this fuzzy area, statements can be both true and false – for instance, they can be 60% true and 40% false. Yet, fuzzy logic is also a form of binary logic – one that posits oppositions as neither 100% true nor 100% false. As Giangiacomo Gerla remarks, ‘a fuzzy logic cannot be an alternative to classical logic since it is a construct of this logic and, at the same time, fuzzy logic is an attempt to extend classical logic’. See Gerla 2017, p. 442.
Alternatives to the CI have been formulated both in the West and in the former Soviet Union (see Cross 1991; Kojevnikov 2013). For example, Blokhintsev accepted the Copenhagen view in 1944 but rejected it in 1949, arguing instead that the wave function provided an objective description of an objective phenomenon (Cross 1991, p. 740). Fock ‘argued that the wave-function described observer-independent “potential possibilities”, one of which was “actualized” as an objective state when a measurement was made. Thus he took the controversial step of extending materialism to potential situations’ (Cross 1991, p. 741). Don Howard argues that ‘the Copenhagen interpretation is an invention of the mid-1950s, for which Heisenberg is chiefly responsible’ (Howard 2014, p. 669). An aspect closely related to superposition is subjectivism. Popper remarks that ‘objective reality has evaporated’ in the CI and that ‘quantum mechanics does not represent particles, but rather our knowledge, our observations, or our consciousness of particles’ (quoted in Howard 2014, p. 679).
Ghirardi and Bassi 2020.
As is well known, Einstein did not accept the notion of superposition.
Marx 1983, p. 86; emphasis added. For a more extended assessment, see Carchedi 2008.
The account in Ananthaswamy 2023 is recommended.
Ball 2021.
Wendt 2015, p. 67.
de Ronde and Massri 2019, p. 7.
Ball 2021; emphasis added.
de Ronde, Domenech and Freytes, n.d.
Ibid.
Bohr, Mottelson and Ulfbeck, p. 15.
See Chapter 2 of Carroll 2019.
Frank and Siegel 2022.
Daniel 1989, p. 255.
Forman 1971, pp. 3–4; emphasis added.
The most noticeable exception was Einstein.
Forman 1971, p. 39.
Ibid.
Forman 1971, p. 101.
‘Einstein was convinced, and rightly so, that his fellow physicists were rushing to embrace a failure of causality without having made any serious attempt to explore the possibilities for a causal solution’ (Forman 1971, p. 95).
For an account of the dialectics of individual and social knowledge and of why some individuals become carriers of pro-capital- and others of pro-labour forms of knowledge, see Carchedi 2011 and Carchedi 2022, second section.
Hassani 2020. The author does not specify what this unique feature is. It is probably the co-existence of irrationalism and positivism.
Hassani 2020.
Hassani 2020; emphasis in the original.
Lukács 2021, p. 214.
Lukács 2021, p. 222.
By the same token, one can hold that we may observe and even criticise the phenomena associated with capitalism but that we cannot explain their essence, which is capitalism itself.
Lukács 2021, p. 9.
To mention just a few of the perceived threats to capitalism: Germany, 1918–20; Hungary, 1919; Italy, 1918–20; the Third International, 1919.
Lukács 2021.
See note 56, above.
For a more general critique, see Carchedi in Wright et al. 1989, pp. 105–25.
See Carchedi 2011.
Lukács speaks of the ‘irrationalist upsurge in imperialist times’. See Lukács 2021, p. 16.