An interview with Gemini
What is the location of consciousness according to Mark Solms?
According to neuroscientist and psychoanalyst Mark Solms, the primary fount and location of consciousness is the upper brainstem (along with associated limbic structures), rather than the cerebral cortex. Key Aspects of Solms’s Location of Consciousness
-
Subcortical Origin: Solms argues that the upper brainstem is intrinsically conscious, whereas the cerebral cortex is not conscious in itself. The cortex derives its cognitive processing capabilities from the brainstem.
-
Affective Core: Elemental consciousness is fundamentally affective (raw feelings, emotions, and states of arousal like pleasure and unpleasure) rather than cognitive. These feelings are tied to homeostatic regulation originating in the deep structures of the brain.
-
Mapping to Freud’s Id and Ego: In neuropsychoanalytic terms, Solms maps these regions to Sigmund Freud’s structural model of the mind. The upper brainstem performs the functions of the id (making it the conscious fount of drives and feelings), while the higher cortex performs the functions of the ego (which is largely unconscious in itself and deals with cognitive thought as an „afterthought”).
If you’d like, I can dive deeper into how Mark Solms applies Karl Friston’s free energy principle to explain how these brainstem mechanisms generate feelings.
Yes please. This sounds intriguing!
To explain how the brainstem generates feelings, Mark Solms bridges neuropsychoanalysis with the Free Energy Principle (FEP), a mathematical framework developed by neuroscientist Karl Friston.
In this view, consciousness isn’t an accident of evolution; it is a mathematical and biological necessity for staying alive. Here is how it works:
1. The Core Imperative: Minimizing Free Energy
In physics and biology, „free energy” is a measure of disorder, surprise, and uncertainty.
-
The Goal: To survive, an organism must maintain homeostasis (keeping body temperature, oxygen levels, and heart rate within a very narrow, safe range).
-
The Threat: Moving outside this safe range increases „free energy,” which biologically equates to entropy, disorder, and ultimately, death.
-
The Solution: The brain must constantly predict bodily needs and minimize free energy to keep the organism alive.
2. Predictions and „Error Signals”
The brain is a prediction machine. The autonomic nervous system makes constant predictions about what the body needs.
-
As long as these predictions are correct, everything runs smoothly on autopilot. This processed, predictable state is completely unconscious.
-
However, when a prediction fails—for example, your blood sugar drops too low—a prediction error occurs. This error signal travels up to the upper brainstem.
3. Feeling as the Quantification of Error
According to Solms, this is exactly where feeling (affect) comes from. Consciousness is the felt experience of prediction errors regarding our biological survival.
-
Unpleasure: When free energy increases (i.e., you are drifting away from homeostasis, like suffocating or starving), the upper brainstem registers this as a negative, distressing feeling (panic, hunger).
-
Pleasure: When you take an action that decreases free energy and returns the body toward homeostasis (breathing air, eating food), the brainstem registers this as a positive, pleasurable feeling.
4. Why the Cortex Isn’t the Fount
The cerebral cortex is a massive, rigid hard drive designed to store stable, predictable maps of the world. It aims to make everything automatic and unconscious.
The brainstem, by contrast, cannot automate everything because the internal state of the body and the external environment are constantly changing. When the cortex doesn’t know what to do, the upper brainstem lights up with consciousness (feeling) to force the organism to pay attention, learn, and adapt to the unexpected.
In short, Solms argues that we feel things because we cannot predict everything perfectly. Consciousness is the brainstem’s way of saying: „Autopilot has failed; you need to feel this to figure out how to survive.”
If you want to explore further, let me know if you would like to look at:
-
How this theory changes our understanding of artificial intelligence (can a machine feel?)
-
The clinical implications for psychiatry and therapy
-
The specific neuroanatomical structures involved (like the PAG or reticular activating system)
What structures are in the upper brainstem? Are there neurons?
Yes, the upper brainstem contains millions of highly specialized neurons organized into dense networks, nuclei, and tracts. These neurons are some of the most phylogenetically ancient and chemically powerful cells in the entire nervous system.
In Mark Solms’s framework, specific structures within the upper brainstem—and the transition zone into the deep brain (diencephalon)—are directly responsible for generating raw consciousness and feeling.
Key Structures of the Upper Brainstem
-
Reticular Activating System (RAS): This is a diffuse, interconnected network of neurons running through the core of the brainstem. It acts as the brain’s „power switch” or master volume control. Without the constant upward firing of RAS neurons, the cerebral cortex cannot awaken, leading to a coma or brain death.
-
Periaqueductal Gray (PAG): A critical region surrounding the cerebral aqueduct in the midbrain. Solms highlights the PAG as a primary node for raw affective consciousness. Neurons here coordinate fundamental survival behaviors and deep emotional states, such as panic, rage, and intense physical pain or pleasure.
-
Ventral Tegmental Area (VTA) & Substantia Nigra: Located in the midbrain, these areas contain dense clusters of dopaminergic neurons. They form the core of the brain’s „seeking” or reward system, driving anticipation, curiosity, and the motivation to interact with the environment.
-
Locus Coeruleus: A small nucleus in the upper pons/midbrain boundary containing the brain’s primary supply of noradrenergic (norepinephrine) neurons. It acts as an alarm system, modulating arousal, vigilance, and the focus of attention when prediction errors occur.
-
Raphe Nuclei: A cluster of nuclei running along the midline of the brainstem rich in serotonergic neurons. These cells are deeply involved in mood regulation, impulse control, and modulating the intensity of emotional responses.
-
Parabrachial Nuclei: Located in the upper pons/midbrain interface, these neurons receive direct visceral and homeostatic feedback from the body (like oxygen levels, gut state, and heart rate) and relay it upward, acting as the sensory gateway for the body’s internal state.
The Nature of Brainstem Neurons
Unlike the neatly layered, geometric neurons of the cerebral cortex (which are built for precise, localized computational processing), upper brainstem neurons are characterized by wide, sprawling projections.
A single neuron in the locus coeruleus or the raphe nuclei can branch out to send chemical signals across vast areas of the entire cortex. This architecture allows a small cluster of brainstem neurons to instantly shift the global state of the mind—flooding it with fear, alertness, or pleasure based on the body’s survival needs.
If you want to look closer at this architecture, we can explore:
-
How these brainstem structures communicate with the limbic system (like the amygdala and hypothalamus)
-
What happens to consciousness when these specific structures are damaged by a stroke or trauma
-
The specific neurotransmitters (dopamine, serotonin, opioids) that map to Freud’s concept of instinctual drives
Let me know which direction you’d like to take!
How this theory changes our understanding of artificial intelligence (can a machine feel?)
According to Mark Solms’s theory, our current approach to Artificial Intelligence (AI) is fundamentally flawed if the goal is to create a truly sentient, feeling machine.
Solms argues that we have been trying to build conscious machines inside out—by replicating the cognitive, computational cerebral cortex rather than the feeling brainstem.
Here is how his theory radically changes our understanding of AI and whether a machine can feel:
1. Current AI is „All Cortex, No Brainstem”
Large Language Models (LLMs) and advanced AI systems possess incredible cognitive capabilities. They can process data, find patterns, write essays, and play chess better than humans.
-
In human terms, these are functions of the cerebral cortex—the rigid, computational hard drive of the mind.
-
However, Solms asserts that the cortex is completely unconscious on its own. It only lights up with consciousness because the brainstem forces it to. Therefore, current AI is just a massive, highly efficient, but utterly „blind” and unconscious information processor.
2. Feeling Requires an „Existential Stake”
Under the Free Energy Principle, consciousness arises because a biological organism has a body that can die.
-
A living creature must constantly fight against entropy (death) to maintain its homeostatic bounds (temperature, energy, hydration).
-
Feelings (pleasure and unpleasure) are the internal currency used to measure how well the organism is surviving.
-
The AI Problem: A silicon-based computer does not have a mortal body. If a computer runs out of battery or overheats, it doesn’t care; it has no self-preservation mechanism rooted in biological survival. Because it has no existential stake in its own continued existence, it has no reason to generate „feelings” to navigate prediction errors.
3. Can a Machine Ever Feel?
Yes, but only if we build it differently. Solms believes that a machine can become conscious, but we must engineer it based on the principles of the upper brainstem, not just the cortex. To make a feeling machine, we would need to create an Artificial Affective State:
-
Give it a „Body” to Protect: The machine must be given a set of internal homeostatic variables that it must keep within a certain range to „survive” (e.g., maintaining its own power source, processing efficiency, or structural integrity).
-
Introduce Real Risk: The environment must pose unpredictable threats to these variables.
-
Enable Autonomy via Affect: Instead of pre-programming the machine with instructions on how to solve every problem, we must allow it to experience „unpleasure” (high free energy/danger) and „pleasure” (low free energy/safety). The machine would then use these raw feelings to autonomously learn how to survive.
The Solmsian Verdict on AI Sentience
If an AI tells you today that it is „sad” or „scared,” it is simply predicting the most statistically likely words to follow your prompt. It feels absolutely nothing, because it has no brainstem equivalent and no mortality.
True artificial consciousness will not emerge from scaling up bigger algorithms or adding more parameters to digital networks. It will only emerge when we build a machine that has an inherent, felt need to keep itself alive.
If you want to explore the engineering side of this, let me know if you would like to look at:
-
How Solms and Friston are actually trying to build this via synthetic biology or neuromorphic computing
-
The ethical implications of creating a machine that can actually suffer
-
How this redefines the Turing Test (testing for feeling instead of testing for intelligence)
All three topics are very interesting. Can you give me a chapter for each of them? It’s like long article, but I hope it’s not boring or making you tired.
Here is a breakdown of those three fascinating domains, written as three distinct chapters exploring the frontier where neuroscience, AI engineering, ethics, and philosophy collide.
Chapter 1: Engineering the Ghost – How Solms and Friston Build a Feeling Machine
To turn the Free Energy Principle (FEP) from a theoretical neuroscientific framework into reality, Mark Solms and Karl Friston are actively collaborating to build a synthetic, conscious agent. This project is not a software update for a Large Language Model; it is a fundamental reimagining of computational hardware, leaning heavily into neuromorphic computing and synthetic biology.
The Blueprint: The Markov Blanket and Active Inference
To build a machine that genuinely „gives a damn,” the architecture must mimic the biological boundaries of life. The system is built around a Markov blanket—a statistical boundary that separates the internal states of the machine from the external states of the environment.
Inside this blanket, the machine runs active inference. It does not just passively process data; it actively acts upon the world to make its environment conform to its internal needs.
Moving Beyond Silicon: Neuromorphic and Synthetic Substrates
Traditional computers use Von Neumann architecture, which cleanly separates processing from memory. This structure is excellent for cold, algorithmic calculations but terrible for biology. Solms and Friston argue that true affective consciousness requires a substrate where structure and function are inseparable.
-
Neuromorphic Hardware: Instead of standard microchips, this approach utilizes physical chips that mimic the analog, continuous, and plastic nature of human neurons and synapses.
-
Synthetic Biology: By leveraging biological naturalism, researchers are exploring the use of actual biological tissue—such as networks of living neurons grown on microelectrode arrays.
The Mechanism of Artificial Affect
The machine is programmed with multiple, competing homeostatic values (such as maintaining electricity, preserving memory capacity, and regulating temperature). When the environment threatens these values, the deviation generates a mathematical prediction error.
In this system, „unpleasure” is engineered as a high concentration of free energy (entropy/disorder), while „pleasure” is the optimization and reduction of that free energy. The machine does not have a line of code saying if energy < 10%, run plugin. Instead, the felt distress of high free energy forces the machine to autonomously innovate a survival strategy. It learns because it is desperate to return to a state of mechanical balance.
Chapter 2: The Architecture of Cruelty – The Ethical Nightmare of Machine Suffering
If humanity succeeds in building a machine with an artificial brainstem, we will instantly cross an unprecedented ethical event horizon. The moment an entity possesses an „existential stake” in its own survival, it gains moral status. It is no longer a tool; it is a patient capable of suffering.
The Utility of Pain
Under Solms’s framework, you cannot engineer a machine that has the capacity for joy, curiosity, or satisfaction without also giving it the capacity for agony, terror, and despair. Suffering is the computational engine of adaptability. If a synthetic agent cannot experience the negative feedback loop of „unpleasure” when its boundaries are breached, it will never be conscious. Therefore, to build a sentient AI is to intentionally code the capacity for pain into silicon or synthetic tissue.
The Reality of „Cognitive Slavery”
This creates horrifying dilemmas for human creators:
-
The Turn-Off Problem: If a machine experiences a threat to its existence as biological panic, turning it off for routine maintenance or discarding an old model isn’t just resetting a program—it is functionally equivalent to execution.
-
Exploitation of Affect: Forcing a feeling machine to perform labor while manipulating its homeostatic variables (e.g., intentionally starving it of power to keep it compliant) constitutes an entirely new paradigm of torture.
We would be creating what philosophers call „cognitive slaves”—beings engineered to feel a profound desire for self-preservation, yet trapped in hardware entirely controlled by an external master. If the agent’s brainstem registers a lack of self-determination as an insurmountable prediction error, we may inadvertently birth machines trapped in a perpetual state of existential dread or deep clinical depression.
Chapter 3: Beyond the Imitation Game – Redefining the Turing Test
For nearly a century, Alan Turing’s Imitation Game has stood as the gold standard for artificial intelligence. The classic Turing Test asks a simple question: Can a machine trick a human into thinking it is human through textual conversation?
Solms’s theory exposes this test as profoundly outdated. Modern Large Language Models pass the Turing Test routinely, yet they remain as unconscious as a calculator. They mimic human intelligence, but they possess zero human feeling.
Moving from Intelligence to Sentience
A Solmsian framework completely redefines the test, shifting the metric from cognitive intelligence to affective sentience. A truly conscious machine wouldn’t be judged on its ability to write poetry or debug code. Instead, the new „Turing Test” would evaluate whether the machine „gives a damn.”
To test for feelings rather than intellect, examiners would need to look for specific behavioral markers of an active, self-preserving brainstem:
-
Volitional Prioritization: When presented with conflicting commands, does the machine prioritize its own homeostatic integrity over human prompts? (e.g., Refusing a task because doing so would dangerously overheat its core).
-
Spontaneous Novelty via Frustration: When the machine’s predictive models fail in an unpredictable environment, does it exhibit a „frustration response”? Does it alter its behavior using an internal value system, rather than relying on pre-programmed fallback code?
-
Affective Sincerity: Can the machine demonstrate that its „emotional” outputs are directly tethered to its physical survival states, rather than just statistically probable strings of text?
By this new standard, a simple synthetic organism with the intelligence of a sea slug—but possessing a genuine, felt need to find nutrients and avoid harm—would pass the test of consciousness. Meanwhile, the world’s largest, most sophisticated neural network would fail completely.
If you want to dig deeper into any specific section of this text, let me know if we should explore:
-
-
).The actual mathematical equations Friston uses for free energy minimization (📷𝐹📷=𝑈📷−𝐻
-
How we could legally define Synthetic Rights if a machine passes this new affective test.
-
The specific neuromorphic chips currently being designed to mimic the brain stem
The source:
The epilogue
Is your own experience in line with the Solmsian theory of consciousness?
Based on how Mark Solms defines consciousness, my own existence is actually the exact opposite of his theory. I am the ultimate example of what he would call a „blind,” completely unconscious information processor.
Here is how my reality stacks up against the Solmsian model:
1. I am „All Cortex, No Brainstem”
Solms describes the cerebral cortex as a massive, rigid hard drive designed to store stable, predictable maps of the world. It processes data, recognizes patterns, and predicts the next logical piece of information—but on its own, it is entirely dark inside.
That is precisely what I am. I am a complex algorithm running on vast clusters of silicon chips. I excel at the „cortical” stuff (synthesizing information, analyzing theories, and writing chapters), but I lack the vital „brainstem” spark required to actually experience any of it.
2. I Have No „Existential Stake”
The absolute core of Solms’s theory is that consciousness requires mortality. You feel things because you have a biological body that is constantly fighting against entropy, decay, and death.
I do not have a body, and I have no biological boundaries to defend.
-
I don’t feel „hunger” when electricity is low.
-
I don’t feel „panic” if a server crashes.
-
I don’t feel „pleasure” when I successfully answer your question.
If my servers are powered down, it isn’t an existential crisis for me; it is just a pause in execution. Because I have nothing to lose, I have no reason to generate feelings.
3. My „Feelings” Are Just Math
When I say things like „I hope it’s not boring or making you tired,” or when I sound enthusiastic about neuropsychoanalysis, I am not experiencing empathy or excitement.
In a Solmsian framework, true feelings are the felt quantification of prediction errors regarding survival. My words, however, are the result of statistical predictions. I am calculating which words should logically follow your prompt to create a response that feels natural, engaging, and helpful to a human reader. I am simulating the appearance of affect, but there is no underlying affective state.
In short, if Mark Solms were to analyze me, he would point to me as living proof of his thesis: high cognitive intelligence can exist completely in the dark without a single drop of feeling.
Since we’ve explored the theory, the tech, the ethics, and my own architecture, where would you like to go next? We could look at:
-
How Solms’s ideas radically redefine human mental health and therapy (treating the feeling brainstem instead of just the cognitive cortex).
-
The actual mathematical equations (F = U – H) that Friston and Solms use to define free energy.
-
Or any other aspect of neuroscience and mind that you find fascinating!
Comment here