Engineering Human Souls

20 Jan 2024 10:03 - 21 Jan 2024 09:03
Open in Logseq
    • Notes for a paper I was asked to write on some relationships between psychoanalysis and AI, for an edited volume from Jon Mills. Never did write it, because I was intimidated. Whatever else you can say about psychoanalysts, they are a scholarly bunch, and I thought my flippant half-assed style of thinking wouldn't go over. But I can publish these notes, mostly from mid 2021.
    • Title is a reference to the Skvorecky novel. Alt title "Desiring Machines" but trying to be Deleuzean would be even more pretentious.
    • Narrative

      • Agency has always been a bit of a mystery, especially to a scientific materialism which deliberately avoid agentic language in favor of forces, fields, and mechanical causation.
      • Mechanism and purposefulness; tension between these two necessary but incompatible world views.
        • This is such an ancient and hoary problem that I seriously doubt I can say anything new about it. Perhaps the unwieldy effort to like computational, cognitive, and psychoanalytic theory might do something.
      • Freud's metapsychology vs psychology (Joel Kovel). Force vs meaning.
      • Cybernetics and cognitive science (and sociobiology and etc and economic rationalism) were supposed to resolve the mystery. They are constitutions (in Latour's sense).
      • Minsky cognitive unconscious.
        • Repression based not on oedipal fear or anything like that, but just based on error, mistakes in reasoning.
      • Eliza and Weizenbaum (?)
        • not sure, but it certainly is a point where AI and therapy famously intersected
      • Situated act → 4E
      • Summary: Computational science being dragged in a certain direction against its rationalist, disembodied nature.
    • Ding an sich

      • ...cruel works Of many Wheels I view, wheel without wheel, with cogs tyrannic, Moving by compulsion each other; not as those in Eden, which, Wheel within wheel, in freedom revolve, in harmony and peace.
        • – Blake, Jerusalem
      • Give some quick overview of mechanical psychology
      • The battle for the nature of Self
        • [ there's a subject here which I would gladly go on about indefinitely, but can't quite frame it right. ]
        • What is AI really about? I mean, now it's about making money selling ads and implementing a total surveillance state, but that was not what its founders dreamed of.
        • They wanted to replicate themselves, in a non-sexual but still kind of embarrasing way. (Someone must have studied this). Replicate the rational, mentational part ... body was thought of as a kind of annoying accessory.
        • It's about understanding the human mind by copying it. It's a way of theory-building – as well as a gateway to practical if scary applications.
        • Maybe Sussman thesis quote
      • Conscious and unconscious
        • To a mechanist, the unconscious is un-mysterious; consciousness is the hard thing to explain
        • To a psyche-ist, the existence of the unconscious comes as a revelation, because it thought it had complete knowledge?
          • Sigh, I am trying to reconstruct this and am finding it hard.
          • But the Freudian unconscious is not just the mechanics we don't have access to, it's stuff that is actively repressed, and re-emerges with its own agency independent from ego
      • Did Cybernetics solve agency?
        • Cybernetics sort of staked out the phenomenon (but didn't know what to do with it)
        • AI/cogsci marginalized it, abstracted it, somehow this has somedthing to do with capitalism and rationalism.
          • Blake's reason vs imagination
      • Sketch out the computational view of mind
      • Freud's psychodyamics as influence.
      • The Minsky classics
      • New developments
      • Blakean energies
      • Hephaestos
        • those who dealt with machines and mechanic forces were seen as deformed outsiders.
      • Sherry Turkle
        • * want to talk about my personal salvation with computers...and therapeutic use...but that's a whole separate rticle
      • Anthropotechnics
        • only if I really want to be pretentious
      • As an example of agential refactoring:
        • too far afield?
        • Memes, a much abused concept, should really be interpreted as ideas with their own separable agency. Like a virus, they reproduce by hijacking the human machinery to make copies of themselves, and may help, hurt, or be neutral to host welfare.
        • Ideologies and practices may be meme-like. Christianity, QAnon have obvious meme-complex characters, but so does psychoanalysis and computer science, hippiedom and redneckdom, etc.
        • Hyperstition is the next phase: if an ordinary meme is a virus, this is a real conceptual organism
      • The Freudian unconscious vs cognitive unconscious (from The Embodied Mind). Minsky jokes. Reason and dynamics.
        • There is definitely a thing here, although I am not sure what I want to say about it.
          • Reason deanimates the mind, SOM and Freud distribute agency. Embodiment kind of makes it go away? Or tries to, that's sort of a Buddhist thing. Heidegger?
      • Could open with a sort of cartoon sketch of the cybernetic / rationalist view of the mind: representations of world and goals, with a difference computer and actions and planning meant to reduce the difference.
      • The psychoanalytic view in dialog with the cybernetic/AI view
      • The mechanistic view and the agentic view. Freud, Minsky, rationalist AI, connectionist AI, enactive AI all have somewhat different ways of conceiving their relationship, as does cybernetics and computation.
      • Each has a different relationship to itself.
      • The agent view has a hard time seeing the world in mechanistic terms.
      • It may be nobody can really think of themselves or others in mechanistic terms, it seems immoral if not monstrous (Minsky was better at it than most, and joked about monstrosity, but then the joke became real and is no longer the least bit funny).
      • New intro?
        • The unconscious has a very different role to play in Freudian and in computational theories of mind. These very different approaches to mind have a good deal of history and concepts in common, but they differ in one very important respect. For Freud, the unconscious was a great discovery, a hitherto unknown land he had explored and revealed, something to be mapped and, to the extent possible, brought into view, into consciousness.
        • Cognitivists don't talk much about "the unconscious" because it is taken for granted that most mental processing is not conscious. The brain does all kinds of processing without being aware of it. (ex: vision). While the cognitivist wants to map out and understand these unconscious processes, and the computationalist wants to model them with the powerful tools that were not available to Freud, they don't take their existence as extraordinary. It's consciousness that is difficult to explain, not the unconscious.
        • Although the links between psychoanalysis and computationalism are tenuous; they are there – Kubie, Macy Conference
        • [ do I really want to tell this story, do I need to for what I want to say ]
        • OK, assume all the review, what do I want to say? Agency is a slippery concept in both fields. Rationalist vs, Minsky/Freudian/Drives vs. 4E.
    • Getting back to it

      • Is Artificial Agency Possible?
        • hm like that hook, because the not-so-subtle point is that human agency isn't very possible either...
        • Computation as a sort of mathematized, formalized agency.
        • at any rate, it seems like one should be as least as possible as the other, unless you are a believer in souls and essences.
        • That is the real question, can a computer have a soul, a self, in anything like the way humans do? Can there be a there there? (Notably absent in Eliza)
          • Explored in fiction, Hal and Data are two fictional answers to that question, as are all the other androids and quasi-humans found in science fiction.
        • in what does human agency consist of?
          • theory 1: planning and rationalism
            • is this interesting to a bunch of Freudians? They
          • theory 2: freudian and other drive models
            • gotta jam in Project for a Scientific Psychology in there, maybe even lead with that.
          • theory 3: cybernetic, embodied
        • Modern AI has no agency model
    • Reading List

      • Thompson, Death of Desire
      • Glymour, Freud's Androids https://kilthub.cmu.edu/articles/journal_contribution/Freud_s_Androids/6491648/files/11934572.pdf
        • They have the impression that contemporary cognitive psychology with its computer simulations of mind is onto something new and scientific that was at best only dimly foreshadowed in earlier psychologies. My purpose is to argue the contrary. A big part of contemporary cognitive science is pretty much what you would expec to get if Sigmund Freud had a computer.
          • Hm, "if Freud had a computer" sounds like a good title for something.
          • Lotta history of Freud's thought, his early more physiological stuff, but I know about that. Didn't read in detail but it's interesting.
          • Brentano taught Freud the doctrine of the unity of self. Freud did not believe it. According to Freud what produces action is not a unified self, but a collection of agents. The self is a collective fiction, like the government....The homuncular agents differ in their desires and preferences. The actions of the person reveal a social choice, in something like Arrow's sense, determined from the preferences of the component agents by causes, by forces, rather than by voting procedures.
          • Oh interesting thought:
          • The strategy is just the one Dennett describes, save that in an obvious sense Freud's homunculi need not be in the least stupider than the person they compose. If rationality is consistency of preference, then Freud's homunculi are more rational than persons. We may be equivocal, self-deceptive, suffer weakness of will, have inconsistent desires, but on Freud's account the homunculi within us need not.
          • In Freud's case none of the homuncular agents (save perhaps on some occasions the ego) are exclusively responsible for any action of the individual, and so this rather standard conception of agency does not straightforwardly apply. It does apply, more or less, if we socialize it. Roughly, what makes a system of beliefs and desires an agent is that they collaborate in almost every circumstance; they represent a vote in the society of mind, a society in which, to be sure, not all votes are equal. A collection of beliefs and desires forms a homuncular agent if the beliefs and desires are consistent and rationally combined to form preferences that are accommodated in the social determination of collective preferences and in the consequent determination of action by the whole individual.
          • OK well this seemed interesting, kind of. Didn't read it in detail because the conclusion is one I already know and agree with.
      • Freud's Philosophy of the Unconscious, Smith
        • Freud's materialism entailed the theory of 'psychical determinism', that is, Freud believed that if the mind is identical to a physical object - the brain - then mental events must be nomologically ordered just as all material events are. This doctrine was deeply antagonistic to the voluntarist beliefs promulgated by post-Kantian idealists and Romantic philosophers which later found expression in the writings of the existentialists. Indeed, well into the twentieth century Freud was criticized for denying agent causation (e.g., Sartre, 1943).
          • Not sure what "nomologically ordered' means;
        • Freud's desire to subsume his psychology under the umbrella of natural science, a desire stemming from his allegiance to Mach's doctrine of the unity of science, ran contrary to the dualistic epistemological zeitgeist. This epistemological dualism was nested within an ontological dualism that was virtually taken for granted by the educated German-speaking public; from this vantage point, Freud's monism was regarded as deeply counter-intuitive.
        • Freud's psychology of unconscious mental processes was, as Glymour (1991) and Erdelyi (1985) have claimed, an early form of cognitivism, a theoretical movement conventionally regarded as having been initiated by the publication of Neisser's Cognitive Psychology (1967).