The octopus knows something we do not
Dear Reader,
If you want to understand what an alien intelligence might feel like, you do not need to wait for extraterrestrials. You need only look into a tank at a marine research facility. The octopus is not from another planet, but it might as well be. Its lineage split from ours over five hundred million years ago. Everything about how it has evolved to be intelligent is different from how we did. Its nervous system is distributed across its body in ways ours is not. Its sensory world is dominated by inputs we barely have. And when it looks at us, from behind those strange horizontal-slit pupils, there is something behind the look that has not read our theories of mind and does not care about them.
I have been thinking about the octopus a lot lately, because it is the closest thing we have on this planet to a natural experiment in what consciousness looks like when it evolves along a different track. If you want to think clearly about whether machines could be conscious in ways very different from ours, the octopus is the case to study. Not because it will settle the question. Because it will show you how badly your intuitions can mislead you when the subject is not shaped like you.
The basic facts about octopus cognition are stranger than most people realize. Let me lay them out.
An octopus has around five hundred million neurons, which is comparable to a dog. But two thirds of these neurons are not in the central brain. They are distributed through the eight arms, with a substantial concentration in each one. Each arm has its own local nervous system that can process sensory information, make decisions, and act semi-independently. When an octopus reaches out to explore something with one arm, the arm is not just carrying out commands from central control. It is, in some functional sense, doing its own investigation.
This is not a metaphor. Experiments have shown that a severed octopus arm, still alive briefly after being separated from the body, will respond appropriately to touch. It will reach toward food, pull away from harmful stimuli, coordinate its own movements. The arm is not conscious in the way we think about the term. But it is doing something a lot of the cognition our brains handle centrally. Distributed processing, in the octopus, is not a design choice. It is what the animal is.
The implications for consciousness are strange. If the arm is doing something like perception and something like decision-making, is there anything it is like to be the arm? Is the arm part of the same experiential subject as the central brain, or is it a separate experiential subject sharing a body? What happens when the arms disagree about what to do, which they sometimes visibly appear to?
Nobody knows. The honest answer is that we do not have the conceptual tools to describe what is going on. Our theories of consciousness were built for centralized nervous systems with clearly localized minds. Octopus cognition does not fit those theories cleanly.
The sensory world of the octopus is also almost incomprehensible from a human perspective.
Octopuses can see, but in a strange way. They are colorblind in the standard sense, having only one type of photoreceptor. And yet they can change color to match complex environments, matching not just the color but the pattern and texture of the surroundings. How they do this without being able to perceive color has been a mystery for decades.
The current best hypothesis is that they see color through their skin. Octopus skin contains opsins, the same light-sensitive proteins that are used in eyes. These skin opsins can respond to different wavelengths of light. The animal may be perceiving color across its entire body surface, in a distributed way, without needing to route the information through the central visual system.
If this is right, and the evidence is growing that it is, then the octopus has a mode of perception that has no human analog. Not just a different sensory apparatus, but a fundamentally different geometry of perception. We see through two eyes into a unified visual field. The octopus, potentially, sees through its whole body into a distributed field. What that would feel like from the inside is genuinely impossible for us to imagine. We do not have the concepts.
This is the point where analogies with human experience start to break down entirely. Nagel asked what it is like to be a bat. The bat, at least, has centralized neural processing that maps onto our vocabulary of perception. The octopus does not. Asking what it is like to be an octopus is not just asking about different sensory modalities. It is asking about a fundamentally different structure of experience, if experience is even the right word.
There are specific behavioral observations that make it hard to deny the octopus is doing something we would recognize as cognition.
Octopuses solve novel problems. They open jars, escape from complex enclosures, disassemble puzzles they have never seen before. This is not fixed action pattern behavior. It is genuine problem-solving in real time, adapted to the specific situation.
Octopuses use tools. They collect coconut shells, carry them in their arms while moving, and use them as portable shelters. This is behavior that involves anticipating future needs, planning, and coordinating between arms in ways that only make sense if there is some centralized modeling happening even in an otherwise distributed system.
Octopuses appear to recognize individual humans. Different octopuses will treat the same person differently based on prior interactions. Some are shy with certain researchers and social with others. This is consistent with a form of individual recognition and long-term memory that is not obviously present in most invertebrates.
Octopuses play. They will engage in behavior that has no apparent function related to feeding, reproduction, or survival. They will manipulate objects for what looks like the sake of doing it, testing them in various ways, sometimes appearing to be entertained. Play in vertebrates is often taken as evidence of relatively sophisticated cognition. In octopuses it appears clearly present, in a lineage that split from vertebrates before the evolution of the vertebrate brain.
Octopuses dream, or something that looks like dreaming. Recent studies have found that octopuses cycle through different states while sleeping, including states in which their skin patterns change rapidly in complex ways, sometimes matching patterns they have used while hunting or threatened during waking life. This could be a form of memory consolidation, motor practice, or something we have no name for. It looks like the invertebrate equivalent of REM sleep. Whether it involves anything like the phenomenal quality of human dreams is unknown.
Add all of this up, and you have an animal that does most of what we take to be behavioral markers of consciousness. It solves problems, uses tools, recognizes individuals, plays, and dreams. And it does all of this with a nervous system organized on principles almost entirely different from ours.
Let me tell you about my own encounter with the question.
Some years ago I was on a diving trip in the tropics and encountered an octopus in a reef. I was maybe two meters away. It was in a crevice, mostly hidden, watching me with one eye. I stayed still. It stayed still. This went on for several minutes.
At some point the octopus extended one arm. Slowly, deliberately, in my direction. It did not come out of the crevice. It just reached toward me. When the arm was maybe half a meter from my hand, it stopped. Then it slowly retracted the arm, and continued to watch me for another minute or so before disappearing into the crevice completely.
I do not know what happened during those few minutes. I know what I felt. I felt that I was being investigated by something. Something with a perspective, with curiosity, with some kind of assessment happening. This may have been my projection. It probably was, at least in part. But the sense of encounter was unmistakable. Something behind the eye was looking back.
The reason I bring this up is that the encounter with the octopus felt more genuinely alien than the encounter with any AI system I have used. AI systems are trained on human text and produce output shaped by human intentions. When they seem to look back at me, that looking is filtered through everything I know about how they were made. The octopus was not filtered. It was doing whatever it was doing on its own terms, having evolved independently for five hundred million years. If there was anything happening behind that eye, it was happening in a mode I have no way to model.
This is what makes the octopus useful for thinking about machine consciousness. If we cannot know what is happening in the octopus, despite it being a biological being with observable behavior and studied neurology, then our confidence about what is happening in AI systems, with far less known about their internal states, is probably overrated. The honest position, in both cases, is that we do not know. And we should stop pretending otherwise.
There is a specific philosophical point that the octopus makes vivid, and I want to draw it out.
Our theories of consciousness are shaped by the case of human consciousness. We have a centralized brain. We have unified visual experience. We have language. We have a specific relationship between our cognition and our body. All of these shape what we take consciousness to be. When we then ask whether other beings are conscious, we implicitly ask whether they have consciousness of the kind we have.
The octopus reveals that this is the wrong question. If consciousness is a real feature of the world, it can take forms very different from human consciousness. Distributed rather than centralized. Sensorially different rather than analogous. Cognitively different rather than similar. Asking whether an octopus is conscious in the human sense is like asking whether a species from another planet has vision in our sense. They might have something visual, but the specific character of it may be genuinely inaccessible to us.
The right question to ask about the octopus, and by extension about any potentially conscious system, is not whether it has consciousness of the kind we have. It is whether it has any kind of consciousness at all. And if it does, what the specific structure of that consciousness might be, given what we can observe about its nervous system and behavior.
This is a harder question. It has fewer clean answers. But it is the honest question. And it is the question the octopus forces us to confront.
What does the octopus teach us about AI consciousness.
The first thing it teaches is that biological substrate is not required for a certain kind of mystery. The octopus is fully biological, and we still do not know what its experience is like. If AI systems produce behaviors consistent with cognition, the fact that their substrate is silicon rather than carbon is not automatically decisive. We have to look at what the systems are doing, not just what they are made of.
The second thing it teaches is that intelligence can be genuinely distributed. The octopus has cognition happening across its body, with local processing in each arm. Some AI systems, particularly the more advanced ones with mixture-of-experts architectures or agentic frameworks, are also becoming more distributed. If distribution does not disqualify the octopus from potential consciousness, it should not automatically disqualify a distributed AI system either.
The third thing it teaches is that consciousness may not require anything like human self-awareness in the specific form we have it. The octopus probably does not have narrative self-consciousness. It probably does not think of itself as a subject persisting through time. Whatever consciousness it has, if it has any, is presumably very different in structure. This means AI systems that lack human-style self-consciousness may still have some form of experience we would not want to dismiss.
The fourth thing it teaches is humility. We have been studying the octopus intensively for decades and still cannot say with confidence what its inner life is like. Any confident claim about the inner life of an AI system, which we have been studying for far less time and understanding far less of, is running on assumptions we should be more careful about.
I want to close with a specific observation from Peter Godfrey-Smith, whose book Other Minds is the best thing I have read on the octopus.
Godfrey-Smith argues that the octopus represents an experiment in what he calls "an alternative parallel history of the mind." Life on Earth evolved intelligence at least twice, once in the vertebrate lineage that led to us, and once in the cephalopod lineage that led to the octopus. These two experiments started from a common ancestor that was probably a simple worm-like creature with a minimal nervous system. Everything intelligent that has happened since evolved separately on the two branches.
If intelligence can emerge twice, in radically different forms, from a common minimal starting point, then intelligence is presumably not tied to any specific architecture. There is more than one way to build a mind. And if there is more than one way to build a biological mind, there may be many ways to build a mind more generally. Silicon may be one of them. Or it may not. What the octopus tells us is that we should not use our specific case as the definition.
This is a large claim, and I want to be careful not to overstate it. Nothing about the octopus proves that AI systems are conscious. What the octopus does is dismantle the assumption that consciousness must look like ours to be real. Once that assumption is dismantled, the question of AI consciousness becomes a genuinely open empirical question rather than one settled by definitional fiat.
I do not know what is happening behind the eye of the octopus. Neither does anyone else. I want to hold my not-knowing there with the same seriousness I hold my not-knowing about the systems we are building. Both are cases where the honest response is to keep watching, keep learning, and keep asking questions that our current concepts may not be able to answer.
Next month I want to write about crows. Because if the octopus is the case for distributed consciousness in an alien lineage, corvids are the case for something even stranger, consciousness that looks human-like in specific behavioral ways despite being built on very different neural architecture from mammalian brains. What we are learning about corvid cognition in the last decade is shifting how we think about intelligence more generally. Stay with me.
— Transmission Sent —
Niklas Hanitsch
Reference materials
- Peter Godfrey-Smith — Other Minds: The Octopus, the Sea, and the Deep Origins of Consciousness (2016)
- Peter Godfrey-Smith — Metazoa: Animal Life and the Birth of the Mind (2020)
- Sy Montgomery — The Soul of an Octopus (2015)
- Jennifer Mather — Octopus: The Ocean's Intelligent Invertebrate (2010)
- Roger Hanlon — Cephalopod Behaviour (2018)
- https://www.nature.com/articles/s41467-021-24080-1
- https://en.wikipedia.org/wiki/Cephalopod_intelligence
Continue reading
- What is machine consciousness, and does it already exist?
- Crows, ravens, and the intelligence we underestimated
- What is it like to be a bat? Nagel's question in 2026
- Do plants communicate? What the science actually shows
- The Wood Wide Web: What Trees Might Teach Us About Intelligence
Frequently asked questions
How intelligent is an octopus? Octopuses show behavioral markers of intelligence that in vertebrates would be uncontroversial. They solve novel problems, use tools, recognize individual humans, play with objects, and exhibit patterns that look like dreaming. Their neural count is comparable to a dog, and their behavior in many cognitive tasks is comparable too, though the tasks and the architecture behind them are very different.
Are octopuses conscious? Probably, according to most researchers who study them, but nobody can say with certainty. Their nervous system is organized differently from ours, their sensory world is very different, and our theories of consciousness were built primarily for centralized vertebrate brains. Applying those theories to the octopus is possible but the results are always tentative. What can be said is that octopuses meet many of the criteria that in other species are taken as evidence of consciousness.
Why is the octopus so different from other intelligent animals? Octopuses evolved intelligence independently from vertebrates, starting from a common ancestor that lived over 500 million years ago and had only a minimal nervous system. Everything about their cognition developed on a separate evolutionary track. This means their intelligence is not just a variation on the vertebrate theme but a genuinely alternative solution to the problem of being smart.
Can octopuses see color if they are colorblind? Possibly, through their skin. Octopuses have only one type of photoreceptor in their eyes and are colorblind in the standard sense. However, their skin contains light-sensitive proteins similar to those in eyes, distributed across their whole body. The current best hypothesis is that they perceive color through this distributed skin-based sensory system. This is still being actively researched.
What can octopuses teach us about AI consciousness? The octopus is the closest thing we have on Earth to an alien mind. Studying it forces us to abandon the assumption that consciousness must look like human consciousness to be real. If we take octopus consciousness seriously despite our inability to fully understand it, we should probably be more open to the possibility that AI systems may have some form of inner experience that is genuinely alien to us and hard to detect.