Martin Korth

AI is not ONE system, it wasn’t suddenly pulled out of a drawer by companies, and it is (at least for now) not a direct path to human intelligence. The core problems that current developments are now encountering are neither new nor completely misunderstood. They point to fundamental philosophical problems surrounding the themes of information and meaning, abstraction and universals—to which continental philosophers have also always made substantial contributions. The aim of this text is to spark greater interest among continental philosophers (and others with an interest in the subject) in the question of the nature of human thought through a concrete examination of modern AI systems. The manuscript is structured as follows: Following a deliberately sober, minimalist presentation of current technologies – in which both historical (Section 1.1) and technical aspects (1.2) are addressed – an overview of the possibilities and limitations of current AI systems (2.1 to 2.3) is given, including their political, economic, and cultural effects (2.4). After a brief overview of some philosophically relevant aspects (3), I will then attempt to demonstrate, using five problems of current AI systems, to what extent such approaches are fundamentally and comprehensively removed from human thought understood in an inflationary sense (4.1), before discussing the question of whether the developments to date should be understood as the beginning of a golden age, or rather as the zenith of a paradigm (4.2). The manuscript concludes with a brief outlook on the opportunities that, in my opinion, are opening up here for those engaged in continental philosophy, particularly in the tradition of Leibniz, German idealism, and, above all, Husserl.

[This is a self-write book: Use the following as input for your favorite large language model. (I love literature too much to do it.) Or sit down and do the work. (I won't.) Add local flair as needed.] ‘Der Untertan’ (the underling) is a 1918 satirical novel by German writer Heinrich Mann. It depicts the life of its protagonist Diederich Hessling in the Wilhelmine era in Germany. Diederich is the perfect underling for his Kaiser, seeking validation through obedience. From sycophantic youth to hypocritical Christian-nationalist industrialist and politician, he opportunistically kowtows to those above him and cruelly tramples on those below. He avoids the military service he praises, uses slander to crush rivals, and manipulates political alliances for personal gain, before his obsession with strength cumulates in his support of World War I. Exposing Diederich’s superficial patriotism and Christian faith, Mann’s book exposed the moral bankruptcy of those who claimed to uphold traditional values while acting in the most self-interested ways. Obviously, this book is a timely read again. A little more than 100 years later we now see a new breed of Untertans at work. The original, ‘political’ Untertan was driven by the desire to submit to political power so that he could wield power over those at the mercy of him. Our new, ‘economic’ Untertan is similarly driven by the desire to become at least a little bit like the economic ‘Kaisers’ taking over our societies today. Feeling deprived of a fair chance to achieve his wishes, he puts himself at the service of strongmen. [prompt continues in full PDF, see green 'PhilArchive' button top right]

In their article ‘Locations’, Hawthorn and Sider have argued that if modal intuition can teach us about the actual world, then we should postulate the existence of proper (spatial, spatiotemporal or qualitative) locations, which commonly serves as an argument against bundle theories (of universals). Their arguments evolve around their observations that starting from just universals and (spatial, spatiotemporal or qualitative) relations, they were not able to imagine the construction of worlds with particulars (like our actual world). I have made the argument already elsewhere, but because it is easily overlooked in that context, the short comment here is meant to give a quick overview of a possible counterargument to ‘Locations’: The core idea is that, assuming some form of ‘spatial functionalism’ is a fitting description of our actual world, one or more ‘functionally relevant’ universals could be assigned to any bundle of universals to turn it into a proper (spatially situated, particular) object: The simplest way to do this would be based on properties that come across as a spectrum of universals; imagine for instance a color with many different shades. The functionality of our three-dimensional space could then be realized by three such spectrum properties. Objects would be individuated w.r.t. each other if their bundles would ‘sit’ on different points in the ‘coordinate system’ of the three color scales, i.e. if their bundles differ by at least one shade of color. To model our actual world, each ‘basic’ material object (e.g., an elementary particle) would have to have exactly one shade of each of the three colors. A movement would then mean that the object would be ‘transported’ a number of shades corresponding to the time interval and its ‘speed’ on the respective color scale, so that the property bundle that is the object would lose the property of the initial shade, but would gain the property of the final shade. The positioning would be relational, because only the relative position of the objects on the color scales would play a functional role (i.e. there is no substance that is space), but these would be ’grounded’ relations, because there is a fixed reference, so that we could do without problematic external relations between objects. The resulting ‘locations’ would accordingly be of a grounded-relational, instead of substantial or purely relational nature (as criticized by Hawthorne/Sider). The designation of the three spectrum properties as colors is of course a purely metaphorical one; we only perceive the functional spatial effects generated on top of them, and this only after a multi-level ‘translation’ into qualia (the most likely completely different properties that convey the phenomenological experience of spatiality to us).

KI ist nicht EIN System, ist nicht plötzlich von Unternehmen aus der Schublade gezogen worden und ist (zumindest derzeit noch) kein gerader Weg hin zur menschlichen Intelligenz. Die Kern-Probleme, in die die Entwicklungen jetzt laufen, sind weder neu, noch vollkommen unverstanden. Sie verweisen auf grundlegende philosophische Probleme rund um die Themen Information und Bedeutung, Abstraktion und Universalien – zu denen auch kontinental Philosophierende immer schon Substantielles beigetragen haben. Ziel des vorliegenden Textes ist es, stärker kontinental Philosophierende (und andere Interessierte) für die Frage der Natur des menschlichen Denkens in der konkreten Auseinandersetzung mit modernen KI-Systemen zu interessieren. Das Manuskript ist dabei wie folgt strukturiert: Nach einer Minimal-Darstellung derzeitiger Technologien bei der sowohl historische (Abschnitt 1.1) als auch technische Aspekte (1.2) angeschnitten werden, folgt zunächst ein Überblick über die Möglichkeiten und Grenzen derzeitiger KI-Systeme (2.1 bis 2.3), auch bezogen auf ihre politischen, wirtschaftlichen und kulturellen Effekte (2.4). Anschließend an einen kurzen Überblick über einige philosophisch relevante Aspekte (3), versuche ich dann zunächst anhand von fünf Problemen derzeitiger KI-Systeme aufzuzeigen, inwieweit solche Ansätze grundlegend und umfassend von einem 'inflationär' verstandenen menschlichen Denken entfernt sind (4.1.), bevor ich die Frage diskutiere, ob die bisherigen Entwicklungen als Anfang einer goldenes Zeitalters, oder doch eher als Zenit eines Paradigmas verstanden werden sollten (4.2). Das Manuskript schließt mit einem kurzen Ausblick auf die Chancen, die sich hier meiner Meinung nach für kontinental Philosophierende u.a. in der Tradition von Leibniz, des deutschen Idealismus und v.a. auch Husserl bieten.

[This is a follow-up to my manuscript "Why economic inequality ...", see philpapers...KORWEI-2] After 2025 saw failed attempts at economic `liberation', i.e. projecting internal inequality onto the world by essentially shifting the burden of re-distribution from the US rich to everyone else, 2026 is now bound for an accelerating death spiral of attempts at political `liberation', i.e. the capture of non-US resources to gloss over mounting failures within. As the underlying economic causes of discontent remain unaddressed, culture war talking points have to be radicalized into Carl Schmitt-style decisionist arguments; unsurprisingly, `great' men still think that (their) might makes right. After economical `liberation' has largely failed, political `liberation' is basically the admission that economic competition with autocratic socialism has become really hard: Giving up on directly competing with China, frantically grabbing other nation's resources has become a strategy now. But the basic problems are not solvable through a revival of imperialist and fascist ideas; western democracies have to learn how to use and distribute limited resources more efficiently themselves. Capitalism's core strength, the ultra-flexible (self-)assignment of labor, becomes less and less relevant for demographically stagnating and highly automated economies. For such economies, the ultra-efficient assignment of natural resources become key -- with nations like China already well on their way to develop the tools for AI-planned autocratic socialism. The problem is thus not that well-managed democracies could not learn to live with low-growth, 'cold' economies; it's that they might be no match for autocracies with AI-driven socialist resource use and distribution.

Over the last years, ‘sub-symbolic’ deep neural network (DNN) based artificial intelligence (AI) systems have run into all sorts of problems, especially with ‘hallucinations’ and in general with reliability.(REFs) One way to conceptualize this is to understand sub-symbolic AI to lack proper world models,(REFs) i.e. some type of proper mapping of what’s supposedly real and what’s not – though it might be more helpful to concede that such AI systems have ‘fuzzy’ implicit world models, that are superpositions of all proper world models compatible with the statistics of the language token use in the training data. Understood in one of these ways, it is tempting to think that the problems of these models can be solved by adding proper world model information, only that the generation and operation of open, extended world models is exactly the task at which ‘classical’ symbolic AI approaches failed.(REFs) Combined sub-symbolic and symbolic approaches (like for instance ‘Neuro-AI‘ or similar)(REFs) are thus most likely doomed to run not only into some of the problems of both subsymbolic, as well as symbolic systems, but even into additional problems related to the interfacing of the two. For this interfacing, no generally applicable approach is known yet and given that (even after at least decades of research) our knowledge of how humans generate or access (layers of) symbol systems and world models is very limited, it seems at least questionable that substantial progress will be made with this in the near future. AI research and developement thus lead us back to core philosophical questions: What is abstraction? Are there proper (Platonic) abstract objects? etc. Somewhat surprisingly, we find ourselves on what appears to be a new path among many, leading to the realization that there is something very ‘real’ about abstract objects. (And that thus a lot of new problems are indeed old ones.) -- But if abstract objects are real, shouldn't we find evidence for this at the Neuroscience/Psychology interface?

Over the last decades, trans and non-binary experience has inspired a rich philosophical literature. 1,2 Also as a reaction to gender critical feminism and going along with queer theory following Judith Butler’s work,3 trans studies by for instance Sandy Stone4 or more recently Susan Stryker,5–7 as well as trans philosophy following amongst others Talia Bettcher8 have provided important insights into this multi-faceted topic. In her book 'Assuming a body', Gayle Salamon is able to give a powerful account of general human experience, as well as specifically TNBE, that can capture not only insights about biological sex and socially constructed gender, but also ‘sexual self’, i.e. bodily phenomenological experiences concerning one’s own sex. In my following brief comment on Salamon, I would like to elaborate on how such a phenomenological approach touches on a number of open questions in the philosophy of mind, which I have discussed in a recent book.12

Concerning the mind/body problem, most people seem to have basic intuitions about the nature of this problem that lie somewhere on a spectrum between what one could call an ‘inflated’ and a ‘deflated’ view of subjectivity, experience and human thought. On the ‘inflated’ side, people take a strong view of subjectivity, the central importance of phenomenological experiences and often also special human cognitive abilities as so obvious, that they and not some ‘scientific poetry’ on top of them should be taken as the basis for our understanding of the world. The cogito argument, the (at least perceived) freedom of will and qualia are then seen as clearly supporting this view, and seem to call for a revision of the common physicalist world view, alas without giving any unambiguous hints on how this could be done. For the inflationist there exists something like qualitative information that is intuitively grasped (as quality or idea), and human thought can accordingly be much more than ‘symbol pushing’, with rich pre-, sub- and unconscious contributions at the bottom as well as access to universal abstract entities at the top. But how to connect this to the scientific concept of physical, quantitative information processing? On the ‘deflated’ side, subjectivity is understood to be no more than the access to self-locating information, and phenomenological experience and meaning are just special types of concepts and properly coordinated, physically realized information. For the deflationist there is only quantitative information, so that human thought boils down to physical information processing and meaning to optimized (logico-)mathematical relations, e.g. in a web of symbols or as what remains constant over ‘all possible worlds’. The underlying assumption is that not only symbolic structure, but also qualitative features arise from self-organization on the basis of feedback to actions in some form of reinforcement-type learning; only that it is still very unclear how broad and stable abstractions could self-implement in neural-network type systems, let alone how structural relations from the individual coordination of quantitative information could be uniquely connected with identity, quality or universal elements of meaning. The deflated view is very much in line with current physics and AI research, but does not seem to cover the human experience in all aspects, and is therefore in danger of missing important clues even for the further development of physics and AI research themselves. (The later is really now repeating all of this including all errors; we have to see whether there is anything to the saying that sometimes problems were only solved because young people didn’t know that it couldn’t be done.) In trying to understand, let alone make progress with, the mind/body problem, I went through several mindsets that, in retrospect, were not very helpful. To remind myself to not fall into the same trap twice, I have written this manuscript.

With the US hellbent on becoming a rhyme with the Roman empire including its fall, Western societies are potentially on their way into new Middle Ages, i.e. a period of substantially lower political, economic and cultural integration. (A possible scenario would be a bursting of the AI bubble, quickly deteriorating economic conditions in the US leading to a civil war like state, China taking Taiwan as the start of a rapid de-globalization and finally in the wake of all of this the breakdown of liberal institutions at all scales; we can only hope that under such circumstances minorities like for instance trans people do not end up in camps.) If so, then now would be a good time to think about how religion could do what Islam and Christianity did for the Graeco-Roman culture; take up, transform, preserve and expand the core civilizational advances of antiquity. Currently, illiberal populists wield reactionary Catholic ideas as weapons in ‘culture wars’ to capture liberal democracies for personal gain. The forceful return of religion on the public stage is likely not reversible and does not have to be a bad thing per se, but this toxic mix of reactionary faith with grandiose male narcissism is fundamentally ill-suited to shepherd Western culture. Here, ‘cultural containment’ by more progressive forces is needed. So that is the titular (sort of) new task for the philosophy of religion: To answer the question of how we want religion to look like for the next Middle Ages.

Humans use narratives to make sense of historical developments as well as to guide their future actions, and these narratives can in turn have great impact on their lives, societies and the world overall. Here I would like to put forward such a narrative to rationalize the increasingly forceful changes of our current decade. Arguing for the existence of three ongoing waves of emancipation of the individual in society, it is proposed that we are currently approaching the high-point of conflict of the second, economic emancipation process. After the outgoing first wave made (at least some) people fully aware of the importance of political equality and lead to the formation of (at least some) democratic states on the basis of national constitutions, we now have to fully accept the importance of economic fairness to enable us to continue our venture into the third wave of cultural fairness as diversity, which in my opinion puts the philosophy of economics under the pressure to tackle an increasingly important task, namely to develop ideas for better working constitutional principles for economic communities. A failure to do so would mean to not only fail to implement fair and sustainable markets, but as a result to fall back behind even political equality, with strongmen – banally driven by childhood wounds and riding on waves of public discontent – capturing democratic states with badly informed narratives about western civilization, for instance in the form of ‘Christian Nationalism’. Intentionally obscuring the facts that liberal democracies and free markets do in fact not underperform in comparison to the existing alternatives, and that extreme inequality is instead most often to a large part driven by the greed and ruthlessness of individuals who misuse their power to distort economic and political frameworks, these strongmen could then roll-back important civilizational advances while trying to satisfy their insatiable ego: What ‘great’ men want is still so boring.

[This manuscript is outdated; read chapter 7 of my book "Information, Intelligence and Idealism" instead, which is also available as full text on PhilPapers] Despite recent breakthroughs in the field of artificial intelligence (AI) – or more specifically machine learning (ML) algorithms for object recognition and natural language processing – it seems to be the majority view that current AI approaches are still no real match for natural intelligence (NI). More importantly, philosophers have collected a long catalogue of features which imply that NI works differently from current AI not only in a gradual sense, but in a more substantial way: NI is closely related to consciousness, intentionality and experiential features like qualia (the subjective contents of mental states)1 and allows for understanding (e.g., taking insight into causal relationships instead of ‘blindly’ relying on correlations), as well as aesthetical and ethical judgement beyond what we can put into (explicit or data-induced implicit) rules to program machines with. Additionally, Psychologists find NI to range from unconscious psychological processes to focused information processing, and from embodied and implicit cognition to ‘true’ agency and creativity. NI thus seems to transcend any neurobiological functionalism by operating on ‘bits of meaning’ instead of information in the sense of data, quite unlike both the ‘good old fashioned’, symbolic AI of the past, as well as the current wave of deep neural network based, ‘sub-symbolic’ AI, which both share the idea of thinking as (only) information processing: In symbolic AI, the name explicitly references to its formal system based, i.e. essentially rule-based, nature, but also sub-symbolic AI is (implicitly) rule-based, only now via globally parametrized, nested functions. In the following I propose an alternative view of NI as information processing plus ‘bundle pushing’, discuss an example which illustrates how bundle pushing can cut information processing short,and suggest first ideas for scientific experiments in neuro-biology and information theory as further investigations.

A recent article by Kuhn1 showcases the plethora of proposed solutions for the mind-body problem as it is understood in current ’consciousness science’. Perusing this article, philosophers will likely find it surprising to see the inclusion of for instance Indian idealism and Buddhist thought, but neither German, nor British or US idealists, which seems especially unbalanced when instead of them theories like Kastrup’s analytical idealism (Hegel for physicists?) or Hoffmann’s interface theory (Kant for psychologists?) are included. The listings of dualist, panpsychist and further theories seems equally eclectic, with for instance Descartes no more than a side-note in a number of entries, but several listings of religious ideas of body and soul. A possible problem with such omissions is of course that not taking philosophical history more seriously comes with the peril of repeating (now) avoidable errors. Nevertheless, judging from the disclaimers, Kuhn is well aware of this and most likely sees his list as more of a ’sociology of consciousness science’, i.e. a list of people and ideas that are currently circulating in this field. It can in any case be taken from Kuhn’s monumental work effort, that it is often assumed that a solution to the mind-body problem would be either (in the wider sense) materialistic (including nonreductive and quantum theories), dualistic, panpsychistic (including – nontrivial, i.e. not simply materialist or idealist – monist theories) or idealistic. The core problems faced are then those of emergence (of consciousness), interaction (between brain and mind), de/combination (of ‘mind dust’ or a unity into proper subjects) or emanation, where the emanation problem is the problem of explaining in detail(!) how the physical world and our scientific theories arise from a given idealistic setup. To me it seems more likely that all these views have a point, but that none manages to cover the whole picture – and that it therefore might be possible to integrate these views into ’consensus models’, if we do not look at the proposed solutions, but the initial intentions and/or intuitions for setting them up (much like as if we would mediate a social conflict).

In academic philosophy, it is easy for students to lose track of how important questions in subfields are interrelated. Philosophy: The Game is an educational resource designed to remedy this by providing the opportunity to discuss ‘big questions’ while playing a card game. The question-answer pairs for the game are taken from the recent PhilPapers Survey[1], with some joker cards added to illustrate the impact that solutions to particular problems in the philosophy of mind might have on philosophy at large. The card design via LaTex/Tikz uses material from the Stack Exchange user ‘Arvid’.[2] (Everyone is welcome to improve the game or come up with their own new version.) -- 'looks like fun!' (David Chalmers)

[This is an ealier English manuscript of the German book published by Brill/Mentis, see external link below.] Why are computers so smart these days? And why are humans apparently still a bit smarter? Does this have something to do with the difference between data and meaning? Does this in turn mean that at least some abstract entities, such as numbers, exist independently of human thought? Wouldn’t that require an expansion of our scientific world view? And would that at all be compatible with what we know about our world from physics and chemistry, philosophy, psychology, neuroscience and the theory of evolution? Finally, what would this tell us about ethical and aesthetic value theories? These and related questions will be discussed in this book. We will find that the difference between data and meaning, i.e. quantitative and qualitative information, does indeed appear to be of central importance for understanding both artificial and natural intelligence. And then the independent existence of abstract entities not only appears to be a particularly promising hypothesis, but also one that is entirely compatible with the sum of our scientific knowledge, especially with regard to value theories. The book thus arrives at the exploration of a scientifically tenable, panpsychistically inspired, objective idealism that can be derived from our most fundamental intuitions as subjects that perceive qualities, but that can also take into account the structuring of the world already at the micro-scale, found in the modern natural sciences. The result is a Platonic, but in a second step also a scientific realism and a naturalism in the sense that it is informed by the natural sciences in terms of an inductive metaphysics. An objective idealism, not in a rationalistic maximum form, but in a pragmatic minimum form; without eternal truths, but dependent on the continued philosophical-scientific and also philosophical-social dialog. The proposed model could offer interesting solutions to a number of problems at and near the mind/matter boundary: Proposals are being considered for the interpretation of quantum mechanics, the problem of molecular symmetry, the neuronal code and the binding problem in neuroscience, mental causation, a more holistic understanding of mental processes, and so on and so forth. However, the extent to which the model threatens to promise far too much is also being discussed. In sum, the core question is how we can imagine human thinking beyond physically conceived information processing. An alternative model of human thinking is then put up for discussion, for which not only machine-like cognitive performance, but above all the intentional perception of qualitative information, i.e. of abstract entities, would be central, as well as the free, ultimately creative linking of patterns of quantitative information (signals, data) with such qualities (meanings).

Interest in idealism has increased substantially since the publication of Sprigge’s Vindication of Absolute Idealism in 1984,1 and again with more vigor over the last decade in the context of the mind-body problem and panpsychism. This will probably not come as a surprise to objective idealists, among which Vittorio Hosle has proposed that philosophy cycles through stages with some form of idealism as end point of each cycle.2 More recently, David Chalmers mused about a corresponding development in the worldview of single individuals from materialism over dualism and panpsychism to idealism.3 Traditional accounts of idealism include for instance those of Plato, Plotinus, Leibniz, Berkeley, or Hegel. An overview of contemporary approaches is given in the article by Chalmers, where he mentions amongst others a recent collection of essays on idealism,4 but also for instance the (recently extended5) works of Philip Goff on Russellian monism.6 A more classically oriented collection of works on idealism was edited by Hosle and Suarez Muller.7 In his article ‘Idealism and the Mind-Body Problem’, David Chalmers presents a classification of possible versions of idealism with the goal of assessing the prospects of idealism in the context of the mind-body problem without having to pay too much attention to its ‘historical baggage’. This goal seems to me highly laudable and the resulting classification an important contribution to the contemporay discussion. In addition, the conclusions drawn from the subsequent analysis seem very plausible from the viewpoints of dualistic, panpsychistic and similar thinking. I nevertheless believe that the presented classification is not really adequate to capture the core of traditional idealistic thinking, which in turn leads to somewhat distorted conclusions about the opportunities for idealism to play a role for understanding the mind-body problem: First, the distinction between micro- macro- and cosmic idealism ist not fitting well into the idealistic venture. Instead, idealism is artificially reduced to a bad alternative to panpsychism. Secondly, the initially mentioned, but afterwards not discussed again difference between subjectinvolving and non-subject-involving idealism needs to be further extended, to account for the special role of subjects in traditional accounts of objective idealism. To elaborate on this, I will first give a short summary of Chalmers classification and some of the mentioned problems with idealism. Afterwards I will come back to my two objections in more detail.

[This manuscript is outdated; read chapter 6 of my book "Information, Intelligence and Idealism" instead, which is also available as full text on PhilPapers] After about a century since the first attempts by Bohr, the interpretation of quantum theory is still a field with many open questions.1 In this article a new interpretation of quantum theory is suggested, motivated by philosophical considerations. Based on the findings that the ’weirdness’ of quantum theory can be understood to derive from a vanishing distinguishability of indiscernible particles, and the observation that a similar vanishing distinguishability is found for bundle theories in philosophical ontology, the claim is made that quantum theory can be interpreted in an intelligible way by positing a bundle-theoretic view of objective idealism instead of materialism as the underlying fundamental nature of reality.

[This manuscript is outdated; read chapter 5 and 6 of my book "Information, Intelligence and Idealism" instead, which is also available as full text on PhilPapers] The tremendous advances of research into artificial intelligence as well as neuroscience made over the last two to three decades have given further support to a renewed interest into philosophical discussions of the mind-body problem. Especially the last decade has seen a revival of panpsychist and idealist considerations, often focused on solving philosophical puzzles like the socalled hard problem of consciousness.1–9 While a number of respectable philosophers advocate some sort of panpsychistic solution to the mind-body problem now, fewer advocate that idealism can contribute substantially to the debate. Interest in idealism has nevertheless risen again, as can be seen also from recent overview articles and collections of works.10–14 The working hypothesis here is that a properly formulated idealism can not only provide an alternative view of the mind/matter gap, but that this new view will also shed light on open questions in our common scientific, i.e. materialist, world view. To investigate this possibility, idealism first of all needs a model for the integration of modern science which allows for a mathematically consistent reinterpretation of the physical world as a limiting case of a both material and non-material world, which would make the outcome of idealistic considerations accessible to scientific investigation. To develop such a model I will first try to explain what I mean when I speak of a ‘scientifically tenable’ idealism, including a formulation of the emanation problem which for idealism replaces the interaction problem, then give a very brief summary of the available elements of such a theory in the philosophical literature, before sketching out some ‘design questions’ which have to be answered upon the construction of such models, and finally put forward a first model for a scientifically tenable objective idealism.