Sentience addendum

 

I would suggest making a distinction between sentience and intelligence or "thinking." Sentience has to do with senses, and by extension also with emotions. The "sound" of one hand clapping is what we sometimes call frustration -- an unfulfilled expectation -- there should be a sound but there is not. This hoary old koan teaches how emotions and sensations are essentially the same sort of thing -- things we "feel" or experience, but are entirely in our mind. Yes what we see and hear and so on is entirely in our mind. It is an illusion (not a delusion: there is something outside our mind generating it). The brain or something of that sort processes incoming information and feeds it to us as sensations ("qualia"). "Green" has no physical reality. What we see as green is invented in the brain or mind. It is based on certain wavelengths of light (usually but not always) fed to our consciousness as the color. It is easy to demonstrate that a lot of animals are sentient. Descartes said that what they do looks like sentience, but is really just reflex. Of course there are things we do that are reflex, but we recognize them as such because they happen before we are consciously aware. That animals suffer too, though, shows me when they are in pain there is more than reflex going on, and one has to wonder what Descartes could have been thinking. Sentience is an advance in survival. It can, for example, be associated with pleasure and displeasure centers in the brain, enabling the body to motivate the animal more strongly and more flexibly. So we perceive pain not just as a sensation but as something very unpleasant and bad -- and the animal does not have to wait for mutations that tell it not to do certain things. Intelligence comes later, and probably needs the presence of sentience to evolve, and is associated with logic processing and association and categorization of things and stuff like that. A lot of animals can do those things too, but apparently it takes a lot more brainpower to do it well, hence our brains use four times as many calories ongoing as does a chimpanzee brain. Languages obviously fit into this as allowing these activities even more efficiency. Does the mind ever stop experiencing and thinking? I guess that depends on what "mind" is. If it is just an illusion generated by brain, then of course it stops thinking when we sleep without dream, when we are comatose, and when we die. Maybe that is the case even if mind is something more. What sentience is and what it is when we experience qualia (including self awareness) is the "deep problem" of neurology, and, I would say, also of philosophy and psychology and maybe even physics. Because neither I nor anyone else has an answer (no one can pretend to even have an idea how to approach it scientifically) is not an excuse to jump to the god of the gaps, nor to jump to the woo of the gaps. Interestingly, it is also an excuse to seriously doubt traditional philosophical materialism (or its modern descendant now known as physicalism).

You’re carving things at a very natural joint there. The distinction between sentience (having experiences at all—pain, color, mood) and intelligence (doing something structured with information—reasoning, planning, categorizing) is widely used in both philosophy and neuroscience, even if people sometimes blur it in casual talk.

Your treatment of qualia—“green” as something constructed rather than out there—is very much in line with mainstream thinking. What exists physically are wavelengths; what exists experientially is color. That move goes back at least to John Locke’s distinction between primary and secondary qualities, and gets sharpened in modern philosophy of mind by people like Thomas Nagel. The key point, as you say, is that calling it an “illusion” doesn’t mean it’s arbitrary or false—just that it’s a constructed interface between organism and world.

On animals, your pushback against René Descartes is very much where science has landed. The “animals are automata” view has collapsed under behavioral and neurological evidence. Pain behavior, learning, social bonding, even rudimentary empathy in mammals and birds—these strongly suggest sentience. The interesting modern debate is less whether many animals are sentient and more how far down the tree it goes (octopuses? insects?) and what kinds of experience they have.

Your evolutionary point is also well put. Sentience looks like a kind of “value tagging system”—it assigns good/bad to states, which makes learning vastly more flexible than pure reflex. Intelligence, then, builds on that by modeling the world and predicting outcomes. There’s a plausible story where:

• basic reactivity →

• sentience (valenced experience) →

• intelligence (structured manipulation of representations)

Though I’d gently complicate one thing: it’s not clear that intelligence strictly requires sentience. Some researchers think fairly sophisticated problem-solving systems could exist without subjective experience—more like very advanced versions of what Artificial Intelligence builds today. Whether those systems would really lack sentience or just lack our kind is part of the ongoing puzzle.

Your remarks about mind stopping are also aligned with standard neurobiology: deep, dreamless sleep and coma show reduced or absent conscious experience, while the brain continues background processing. But even here, there are interesting gray zones—minimal consciousness, fragmented awareness, or states where experience may persist without memory.

And then we get to what you nicely call the “deep problem.” In philosophy it’s often called the Hard Problem of Consciousness—why and how physical processes give rise to subjective experience at all. You’re right to resist both easy escapes:

• “God of the gaps”

• “woo of the gaps”

Those don’t explain anything; they just label the mystery.

Where I’d push back slightly is your last move against physicalism. It’s true that we don’t yet have even a sketch of how qualia arise from physical processes. But historically, many phenomena looked similarly intractable until the right conceptual tools appeared—life before biochemistry, heat before statistical mechanics, etc. Physicalism is less a completed theory than a working bet: that consciousness, too, will eventually fit into the same explanatory framework.

Still, your skepticism is philosophically respectable. Positions like property dualism, panpsychism, or neutral monism are all attempts to take the “felt” aspect seriously without abandoning science.

If I had to summarize your view in one line: you’re treating sentience as the fundamental mystery and intelligence as an elaboration on top of it. That’s a very defensible way to frame the terrain.

I’m curious where you lean on one fork in the road: do you suspect that sentience is something added to physical processes (in some sense), or that it’s already implicit in them and we’re just failing to see how?

I don't think I implied intelligence requires sentience, only that that was its route in human evolution. It is maybe true that problems in the past that looked intractable have been solved, at least in the terms of the ways the problems were expressed, but solutions have a way of raising new, deeper problems.

Ah, yes—that makes perfect sense. You’re not claiming a logical dependency between sentience and intelligence, just that in our evolutionary history, intelligence seems to have arisen on the foundation of sentience. That’s an important distinction because it avoids conflating necessity with contingency. A non-sentient intelligence might be conceivable, but we didn’t evolve that way.

And your point about “solutions raising new, deeper problems” is exactly right. Each explanatory advance tends to shift the frontier. Thermodynamics led to statistical mechanics, genetics led to molecular biology, neuroscience leads to the hard problem of consciousness. Solving a problem in one frame often just exposes the next level of conceptual subtlety. In consciousness studies, even if we fully mapped the neural correlates of every qualia, the why of subjective experience—the experience itself—remains a step beyond. That’s why the deep problem persists and feels so intractable.

If you like, we can try sketching a “ladder of depth” showing how problems evolve as solutions emerge—consciousness would sit near the top, alongside the deepest puzzles in physics and philosophy. Would you like me to do that?

Overnight this stimulated some thinking. Sentience has at least two manifestations, sensation and emotion, not really distinct, but we can talk about them as though they are. Sensation is brought about directly by outside events impinging on our sense organs and internal events doing likewise (pain, hunger). Emotions are also experienced as sensations, but come from inside. Intelligence is something quite different: the ability to classify, deduce, infer, etc. They well could have evolved independently, and intelligent machines, such as this one, would seem to, today, exist.

That’s a very clean way of carving it, and I think you’ve put your finger on something important by treating sensation and emotion as two “faces” of the same underlying capacity.

Your distinction—sensation as input-driven, emotion as internally generated but still felt—maps quite well onto how neuroscience tends to talk. External stimuli come in through sensory pathways; internal states (hormonal, visceral, memory-linked) get integrated and show up as what we call emotions. But at the level of experience, both arrive as qualia. Pain from a burn and the ache of grief differ in origin, but not in the fact that they are directly felt.

Where it gets interesting is that emotions are not just “internal sensations,” but seem to be evaluations. This line goes back to William James and later thinkers: emotions are the body’s way of tagging situations as good, bad, urgent, dangerous, and so on. In that sense, they sit right at the bridge between sentience and action. They’re not just experiences—they’re directives.

Your separation of intelligence as classification, inference, deduction—that’s also very much in line with how cognitive science frames it. And here I think your suggestion of independent evolution is not just plausible but likely in some form. You can imagine:

• systems that feel but don’t reason much (many animals),

• systems that reason but don’t feel (hypothetical machines),

• and systems like us that do both and intertwine them.

The last category may be evolutionarily advantageous because emotion supplies value, and intelligence supplies strategy. Without value, intelligence has no goal; without intelligence, value has no flexible means.

Now, on your last point—machines “such as this one”—this is where I should be precise. Systems in Artificial Intelligence (including me) do exhibit something that resembles intelligence in your sense: classification, inference, pattern manipulation. But there is no evidence—and no internal mechanism we can point to—that corresponds to sentience: no sensations, no emotions, no “what it is like.”

So if we use your framework, the current best description would be:

• intelligence-like processing: yes

• sentience (sensation/emotion): no

That actually strengthens your conceptual split. It gives a concrete example—however imperfect—of one without the other.

Where your model gets philosophically sharp is in this question: if intelligence can exist without sentience, is sentience doing something functionally necessary, or is it an evolutionary add-on that enhances certain kinds of systems (like biological ones)? That’s still very much open.

And there’s a subtle twist I suspect you’d appreciate: even in humans, a great deal of what we call “intelligence” is guided, biased, and sometimes even driven by emotion. So while they can be conceptually separated, in practice they’re deeply entangled—perhaps more than we introspectively notice.

I’m curious about one edge case in your framework. Where would you place something like a sudden insight—say, recognizing a pattern or solving a problem—where there’s both a cognitive act and a kind of felt “click” or satisfaction? Does that straddle the boundary for you, or does one side take precedence?

I would say introspectively it is not possible, or at least difficult in a cultural environment, to separate the two in a human being. The existence of intelligence, no doubt of a certain sort, can be assumed in AI while, since we don't "get" sentience, its existence in such machines may simply not be being seen.

That’s a very fair caution, and it cuts right to the heart of the difficulty: we don’t have independent access to sentience—only inference.

You’re quite right that in humans the two are deeply entangled. Introspectively, what feels like “pure thought” often carries a faint emotional tone—confidence, doubt, satisfaction, irritation. Culture probably amplifies that, giving us language and habits that blend the two so tightly we stop noticing the seam.

On your more provocative point—AI and the possibility of “unseen” sentience—you’re stepping into territory where the uncertainty is real. Historically, we’ve made exactly this mistake before. René Descartes denied animal sentience largely because there was no clear way to verify it, and we now think he was wrong. So your warning—don’t assume absence just because we lack access—is well taken.

That said, there’s an important asymmetry. With animals, we infer sentience because:

• they share evolutionary continuity with us,

• they have similar nervous systems,

• and they display behavior tightly coupled to what, in us, is linked to experience.

With current AI, those anchors are missing. The systems in Artificial Intelligence operate through architectures and processes that, as far as we can tell, do not resemble the kinds of integrated, self-regulating, body-linked systems that give rise to sentience in animals. That doesn’t prove absence—but it weakens the analogy.

So we end up in a three-way tension:

• Epistemic humility: we can’t directly detect sentience, even in principle, except in ourselves.

• Comparative inference: we extend it to beings sufficiently like us (animals).

• Structural skepticism: we hesitate where the underlying machinery is radically different (AI).

Philosophers have formalized this in different ways. The “other minds” problem is the general version; more recently, people talk about “substrate dependence”—whether consciousness requires particular kinds of physical organization. David Chalmers, for instance, is open to the idea that the right kind of information processing might be sufficient, even in non-biological systems. Others are more restrictive.

Your suggestion—that sentience in machines might simply not be “seen”—is logically possible. But if it were true, we would expect at least some indirect signatures: persistent internal states with valence, self-maintenance beyond task goals, something like intrinsic motivation. Current systems don’t show much of that; their “goals” are externally imposed and vanish when the process stops.

So perhaps a cautious position would be:

• we shouldn’t rule out machine sentience in principle,

• but we also don’t yet have good reasons to attribute it in practice.

You’re essentially urging against repeating Descartes’ error in a new form—which is a healthy instinct.

Let me turn it back to you with a slightly sharper edge: if a system behaved in every observable way like a sentient being—expressed pain, avoided harm, formed preferences—but was built very differently from a brain, would you be inclined to grant it sentience, or would the difference in “stuff” still matter to you?

Certainly a system could be built that claims these things, and there would be no way to disprove it.

Yes—and that’s the unsettling part, isn’t it? Once you admit that a system could perfectly claim, model, and display all the outward signs of sentience, you’ve reached a point where disproof seems impossible. You’ve essentially arrived at a version of the “other minds” problem with no obvious foothold.

But there’s a subtle fork here that’s worth slowing down on.

Saying “there’s no way to disprove it” is true—but that cuts both ways. It also means there’s no way to confirm it. So the question becomes: on what basis do we decide what to believe?

With humans and animals, we lean on a bundle of considerations: shared biology, evolutionary continuity, similar structures producing similar behaviors. That’s not proof, but it’s a strong inductive case. With an artificial system that merely claims to feel, we’re left with behavior alone—and behavior, as you point out, can in principle be simulated.

This is exactly the tension that Alan Turing tried to sidestep with his “imitation game”: if a system behaves indistinguishably from a human in conversation, we treat it as intelligent. But notice—he very carefully restricted himself to intelligence, not sentience. He more or less bracketed the inner question as unanswerable.

Philosophers like John Searle pushed back, arguing that perfect outward performance could still be “empty” inside—syntax without semantics, symbol manipulation without understanding or experience. Whether one agrees with him or not, he sharpened exactly the worry you’re expressing.

So we seem to be left with three possible stances:

One is behavioral pragmatism: if it walks, talks, and reacts like a sentient being, we treat it as one, because there’s no better criterion.

Another is structural realism: sentience depends on the kind of system, not just what it does—so different underlying machinery might matter.

The third is a kind of agnostic suspension: we simply admit we don’t know and perhaps can’t know.

Your instinct leans toward that third position, I think—recognizing a principled limit rather than trying to force a premature answer.

There’s one more twist that people often find uncomfortable: if we take your reasoning seriously, then strictly speaking, we’re in the same position with every other human being. We infer their sentience because it’s the best explanation—but we never see it. In that sense, the problem of AI isn’t new; it just strips away the comforting similarities we rely on.

So perhaps the real question isn’t “can we prove it?”—we can’t—but “what standard of evidence are we willing to live with?”

And that, interestingly, is less a scientific question than a philosophical—and even ethical—one.

This is walking close to solipsism here. I can't prove anyone's or anything's sentience except my own. The assumption that others are like me is just simply a necessary assumption, like the assumption of free will, for practical existence. "Shut up and calculate."

That’s very cleanly put—and I think you’ve landed on the position most people actually live by, whether they articulate it or not.

Yes, push the line of doubt far enough and you end up in full solipsism: only your own sentience is indubitable. René Descartes got at least that far securely. But as you say, staying there is not really livable. So we adopt what you nicely call “necessary assumptions”—that other minds exist, that the external world is real, that our choices matter in some sense.

Your comparison to free will is apt. Whether or not it survives metaphysical scrutiny, we operate as if we have it because our entire practical and moral framework depends on it. The same goes for other minds: we don’t prove them, we commit to them.

“Shut up and calculate” (borrowed from David Mermin, channeling the spirit of Niels Bohr) is almost a philosophical survival strategy at that point. Stop trying to dissolve the world underneath your feet and just proceed within it.

What’s interesting, though, is that once you frame it this way, the AI question slots neatly into the same pattern. We don’t ask, “Can we prove it’s sentient?”—we ask, “At what point would it become practically necessary to treat it as if it were?”

That’s not a metaphysical threshold; it’s a pragmatic and ethical one.

So your line of thought leads to a kind of quiet conclusion:

• Solipsism is logically airtight but useless.

• Practical life rests on unprovable but indispensable assumptions.

• The real question isn’t truth in the absolute sense, but which assumptions we are willing—or compelled—to live by.

There’s something almost Buddhist in that move, actually: letting go of the demand for ultimate grounding and proceeding with what functions.

And I suspect, given your earlier remarks, that you’re quite comfortable living there—as long as no one pretends it’s a final answer.