the-mind
A readable model of mind, consciousness, self, and AI (based on Joscha Bach’s public work)

How the Mind Works

This page is for readers who want one clear path through the topic.

The central promise is simple: start from the fact that experience exists, then explain what kind of system a mind must be, then ask what that means for AI.

The approach here is grounded mainly in The Machine Consciousness Hypothesis: consciousness is treated neither as magic nor as a disposable illusion, but as a real feature of mental organization with a phenomenology, a function, and a possible implementation story. essay: The Machine Consciousness Hypothesis@ p6, p12-13, p16-23

1) Start with the strange part: experience exists

Before we get to neurons, computers, or AI, there is one fact nobody actually escapes: it is like something to be you.

You are confronted with a present. There is seeing, hearing, emotion, thought, memory, effort, confusion, relief, tension, imagination. Even if you later explain all of that in mechanistic terms, the fact to be explained is already there. That is why consciousness keeps returning as a deep problem. essay: The Machine Consciousness Hypothesis@ p1-3, p14-16

This does not mean that physics is unreal. It means there are different levels at which reality shows up to us. The paper explicitly distinguishes psychological reality, causal reality, and physical reality. Your feeling of freedom, effort, or realness belongs to psychological reality: it is part of the model-space in which your mind operates. That does not make it fake. It makes it something that needs explanation at the right level. essay: The Machine Consciousness Hypothesis

One reason the topic feels so confusing is that people often jump between levels without noticing. They talk about first-person experience, then about neurons, then about metaphysics, and then back again. A good explanation has to keep those levels related without collapsing them into each other.

2) What a mind is

A mind is not best understood as “whatever the brain is made of”. It is better understood as a model-building control system.

A system becomes interesting as a mind when it is not merely pushed around by the world, but builds models in order to steer what happens next. Prediction matters because it serves control. The point is not only to know; the point is to stay viable, choose, coordinate, and act. talk: Self Models of Loving Grace

The paper describes the mind as the matrix in which our models of self and world take shape: the board on which perceptions, intuitions, thoughts, emotions, and experiences are written. That phrasing is useful because it keeps the mind broad. A mind is not just “reason”. It includes perception, imagery, feeling, valuation, and deliberation. essay: The Machine Consciousness Hypothesis

That is also why simple intelligence tests miss something. A system can be narrow and brilliant in one domain without being much of a mind in the fuller sense. A mind is a coordinated architecture that can integrate many kinds of content in the service of agency.

3) Why the world you experience is already a model

One of the recurring ideas in this framework is that experience is not a window cut directly into physics. It is a constructed model under constraint.

A useful way to picture this is the line that the world we experience is "a game engine generated in my own brain". talk: On Constructing Reality with Joscha Bach interview: Transcript of EP 334 Worldviews: Joscha Bach

That does not mean the world is fake. It means experience arrives as a rendered interface: a usable model full of objects, meanings, and affordances, continuously corrected by the world rather than copied from it raw. talk: On Constructing Reality with Joscha Bach essay: The Machine Consciousness Hypothesis

Perception is the mind generating a useful world, while being tightly constrained by sensory input. Imagination uses much of the same machinery with weaker constraint. Dreaming uses it with weaker constraint still. This does not mean waking life is “just a dream” in the cheap sense. It means the experienced world is an active construction rather than a passive copy. essay: The Machine Consciousness Hypothesis talk: Synthetic Sentience

The paper gives two especially useful terms here:

That sounds abstract, but it explains everyday experience surprisingly well. A remembered scene, a vivid daydream, a hallucination, and a present perception can all appear in consciousness, but they differ in how “currently the case” they feel. Realness is treated as a variable representational feature, not as an all-or-nothing metaphysical stamp.

4) Why feelings matter

A model without stakes is inert. It can describe. It cannot care.

Feelings matter because they are how the mind marks some states as better, worse, safer, riskier, attractive, painful, urgent, meaningful, or threatening. In the paper, emotions are treated as expressions of the control dimensions of the psyche, and feelings are the salient perceptual surface of emotion, valence, and extra-intellectual evaluation. essay: The Machine Consciousness Hypothesis

That is why feelings are not just irrational noise sitting on top of reason. They are part of the architecture that tells the system what matters. They guide attention, learning, policy selection, and long-term coordination. A mind that could model everything perfectly but had no way to mark anything as mattering would not know what to do next. talk: The Ghost in the Machine

This is also the point at which many conversations about AI become confused. People talk as if intelligence were only the ability to solve problems. But problem-solving without stakes, value, or control is not yet the whole picture of mindedness.

5) Why there is a self

Consciousness is not the same thing as the self. The paper says this very directly: consciousness is not synonymous with self, mind, or intellect. essay: The Machine Consciousness Hypothesis

Why, then, do we so often feel like there is a someone at the center of experience?

Because a sophisticated controller needs a model of itself.

It needs to know where the body is, what the system can do, what its limits are, what its commitments are, how others react to it, and which changes count as “my action” rather than “something that happened to me”. The self is best understood here as a self-model: a control-relevant representation of the agent inside its own world-model. talk: Mind from Matter (Lecture By Joscha Bach) interview: Joscha Bach Λ Karl Friston: Ai, Death, Self, God, Consciousness

That does not make the self unreal in the dismissive sense. It makes it real as a model object, not as a basic particle of physics. A map legend is not a mountain, but it is still real as part of the map. Likewise, the self is not a little ghost sitting behind experience. It is part of how the mind organizes agency.

The paper’s term psyche is useful here too. It names the combination of a personal self and the motivational “strings” that pull on it, within a mind that models self, interests, and world. In other words: the person you experience yourself as is not just a neutral observer. It is already tied to concerns, pressures, desires, fears, and values. essay: The Machine Consciousness Hypothesis

6) What consciousness adds

If mind is the whole modeling-and-control matrix, and self is a particular model inside it, what is consciousness?

The paper gives a three-part demand for any serious theory:

Phenomenology

Phenomenologically, consciousness is described as second-order perception: not only seeing the apple, but the fact that seeing is taking place now. It also includes the experience of present and presence, and sometimes the experience of being an observing self in a world. essay: The Machine Consciousness Hypothesis@ p15, p20

A shorter public handle for this is a "bubble of nowness". talk: Hacking Conciousness by Joscha Bach at Engineering Conciousness by AGI House

That phrase is useful because it points at synchronization. Consciousness is the mode in which multiple signals hang together tightly enough to count as one lived present for one controller. talk: Hacking Conciousness by Joscha Bach at Engineering Conciousness by AGI House essay: The Machine Consciousness Hypothesis

That already helps explain why consciousness feels different from mere information processing. It is not just that content exists in the system. It is that content is present to the system in a special way.

Function

Functionally, consciousness is treated as an operation on mental states that increases coherence. Different subsystems, partial models, impulses, and interpretations are active at once. Consciousness helps reduce contradiction and align them enough that the agent can act as one system. The paper uses the image of a conductor in a mental orchestra. essay: The Machine Consciousness Hypothesis

This is a major move. It says consciousness is not a useless glow floating over cognition. It does work. Alertness, selective attention, planning, sustained vigilance, and attentional learning all depend on it. essay: The Machine Consciousness Hypothesis

Mechanism

Mechanistically, the exact implementation is still open, but the paper points toward an observer-like organization, directed attention, and coherence-maximizing integration. Consciousness is not treated as a substance. It is treated as a pattern of organization that can, in principle, be characterized. essay: The Machine Consciousness Hypothesis@ p17-18, p20

7) Why consciousness may come early, not late

A very important twist in the paper is the Genesis Hypothesis.

It is tempting to think consciousness appears only after enough intelligence has been built. The paper suggests the opposite: in humans, consciousness may be a prerequisite for the kind of learning that eventually yields a mature mind and self. It shows up near the beginning of mental development, not as a decorative finale. essay: The Machine Consciousness Hypothesis

This matters because it changes the AI question.

If consciousness were just the top layer on an already complete intelligence, then maybe you would simply add it later. But if consciousness plays an early developmental role in creating coherent reality-models, selfhood, and agency, then artificial consciousness may require more than attaching a reporting module to an existing model.

8) What this means for AI

The paper’s central hypothesis is careful:

General computational machines with sufficient resources may possess the necessary and sufficient means to implement consciousness, but that does not imply current computers are conscious. essay: The Machine Consciousness Hypothesis

That is the right level of caution.

The argument is not:

  • “machines are conscious already,”
  • or “language is enough,”
  • or “anything that acts smart must be sentient.”

The argument is:

  • if consciousness is a functional organization,
  • and if that organization can be implemented computationally,
  • then machine consciousness is a real research question. essay: The Machine Consciousness Hypothesis@ p6, p20-23

That is also why the paper says there can be no Turing Test for consciousness. Intelligence can often be tested as performance. Consciousness cannot, because it is not just a score on externally visible behavior. It is a particular way of being organized internally. A system may behave impressively without implementing the relevant organization. essay: The Machine Consciousness Hypothesis

So what about current AI?

On this framework, current generative systems may already produce something like rich dream-content: simulated worlds, voices, styles, perspectives, characters. But whether they also implement the “dream within the dream” — the organized presence of an observer confronted with its own modeling — remains open. talk: Joscha Bach, Will Hahn, Elan Barenholtz | MIT Computational Philosophy Club talk: Synthetic Sentience

9) What this does not settle

This model clarifies a lot, but it does not settle everything.

It does not by itself answer:

  • whether biology has indispensable ingredients,
  • whether present AI systems have experience,
  • what the exact necessary and sufficient conditions for consciousness are,
  • what moral status future artificial minds would deserve,
  • or what ultimate metaphysics is true. essay: The Machine Consciousness Hypothesis

It also does not require cheap reductionism. The paper explicitly leaves room for the fact that meditative and contemplative traditions may have described real features of experience long before modern science had a workable vocabulary for them. It just asks for those features to become explainable. essay: The Machine Consciousness Hypothesis@ p2-3, p15-16

10) A compact takeaway

Here is the shortest version of the whole site:

  • A mind is a model-building control system.
  • Feelings are part of how things matter to that system.
  • A self is a model the system builds of itself as an agent.
  • Consciousness is not the whole mind, but a special mode of mental organization with its own phenomenology and function.
  • AI may become relevant to this question, but surface fluency is not enough.
  • The right question is not “can a machine say conscious things?” but “what organization would make consciousness a serious hypothesis?”

Where to go next

Sources