Chapter 4 closed by asking what intelligence actually is. This chapter answers, and it does so by refusing the obvious starting point. Human concerns about intelligence started from the study of reasoning and logic — but is that the right starting point for AI? Is human intelligence the only legitimate reference for us to determine how an artificial intelligent system should be designed, look like, and work? Is there one type of intelligence, or are there many, which we could identify and use as references in the articulated definition of intelligent systems engineering?
The move the deck makes is the same one Chapter 2 made for autonomy: go to living systems. Intelligence is everywhere around us, in the form of intelligent behaviour, emerging as a critical property of living systems in the same way as autonomy does. Living organisms keep on facing ever-changing challenges and survive, exhibiting some form of intelligent behaviour — and they have been doing so since the origin of life on Earth around four billion years ago. A long-term success story, as the deck puts it.
The material of this chapter is drawn from Antonio Damasio’s Feeling and Knowing [Damasio, 2021], with Self Comes to Mind [Damasio, 2010] behind it. It is neuroscience, not computer science, and the deck does not pretend otherwise. What it is doing here is supplying a reference model of intelligence that is not modelled on the rational adult human — which is precisely what an engineering discipline needs if it is not to hard-code one narrow answer into its abstractions.
Four billion years ago, around the end of the intense asteroid impacts on Earth, there were prokaryotic (bacteria-like), single-cell, anaerobic organisms in a very challenging environment — an environment where the appearance of homeostasis provided the most direct path to life maintenance.
These organisms had no feelings, no thought, no reasoning, no mind, no consciousness. And yet they sensed others like them, and sensed their environments. The deck is careful about the word: sensing is not perceiving, not creating a representation, not producing an image for the mind.
On the other hand, sensing is the most elementary sort of cognition. The first living organisms responded intelligently to what they sensed — their response helped their survival. The problem was their potential death; their behavioural response was their intelligent solution to the problem.
Read that last sentence as a definition, because it functions as one. Intelligence, at its root, is a behavioural response that solves the problem of not dying. No representation is required, no reasoning, no language. Everything else — minds, images, deliberation, logic — is a later and optional elaboration on top of this. That reframing is what makes it possible, in Chapter 6, to look at seventy years of AI without assuming that symbolic reasoning was the only legitimate project.
What those first organisms had was no logic, no reasoning, no representation — just an implicit, concealed competence accounting for the goal of maintaining life.
Their non-explicit intelligence was in charge of curating life, managing it in accordance with the rules and regulations of homeostasis — where homeostasis worked as a collection of how-to rules relentlessly executed autonomously according to no explicit directions, yet ensuring that the life-depending parameters (the presence of nutrients, certain levels of temperature or pH) were maintained within survival ranges.
This paragraph answers, in one shot, the question Chapter 1 left open about the conceptual bridge between the two drivers. Homeostasis is simultaneously the elementary form of autonomy and the elementary form of intelligence. Chapter 2 introduced it as “an obvious manifestation of autonomous behaviour”; here it is the substrate of the first intelligent behaviour on Earth. One mechanism, two drivers.
The phrase to hold onto is “how-to rules relentlessly executed according to no explicit directions”. It describes something that has procedural knowledge without having any representation of it — competence without a manual. Chapter 3 already met the computational counterpart of this and gave it a name: the weak notion of agency, where behaviour is directly designed and programmed to achieve a goal which is not explicitly represented.
No metaphysical discussion, no religion, no ethical discourse here — just the simple, innocent perspective of each living organism. Life is inseparable from the apparent goal of maintaining itself, with homeostatically regulated variables paving the way since the early single-cell organisms.
The deck splits that goal in two, and the split is worth keeping because it is the ancestor of a distinction every agent architecture makes:
| Goal | Content | Question the organism is answering |
|---|---|---|
| Internal | Maintaining my own life | Do I have enough nutrients, energy, building materials to perpetuate my own life? |
| External | Saving my own life, avoiding sudden death | Am I in danger, is there any external threat to my life? And if so, how can I counter the threat and survive? |
Two goals, two time horizons, two sources of information — one from inside the boundary, one from outside it. An organism that only watched its internal state would starve without noticing a predator; one that only watched the world would be eaten by its own metabolism. Managing both at once is the original version of the reactive/proactive tension that Chapter 4 said an agent architecture must structurally solve.
The deck gives a timetable [Damasio, 2021], and it is worth studying rather than glancing at, because the gaps between the rows carry the argument.
| Milestone | When |
|---|---|
| protocells | 4 billion years ago |
| first cells, or prokaryotes, without a nucleus | 3.8 billion years |
| photosynthesis | 3.5 billion years |
| first single cells with a nucleus, or eukaryotes | 2 billion years |
| first multicellular organisms | 700–600 million years |
| first nervous cells | 500 million years |
| fish | 500–400 million years |
| plants | 470 million years |
| mammals | 200 million years |
| primates | 75 million years |
| birds | 60 million years |
| hominids | 14–12 million years |
| homo sapiens | 300 thousand years |
Why did nervous systems appear at all? The deck’s answer is strictly functional, and notably unromantic.
The emergence of multi-cellular and multi-system organisms generated a new level of complexity for living systems. Differentiated systems — respiratory, digestive, immune — called for functional coordination. Which was enabled by the appearance and development of nervous systems, improving homeostatic regulation and maintaining life more securely within mindless organisms.
And then the clause that matters most: nervous systems paved the way for mental images, feelings, consciousness, creativity, cultures — as a “secondary” effect.
Minds are a side effect of a coordination problem. The nervous system was not selected for thinking; it was selected because a body made of differentiated subsystems needs its parts kept in register, and it improved homeostatic regulation in organisms that had no minds at all. Everything we usually call intelligence rides on infrastructure built for something else — which is a useful thing to remember when designing a system whose “intelligence layer” is being added on top of a coordination layer.
The deck now walks up the ladder in three steps, and each step adds exactly one thing.
Being. Bacteria have no nervous system and no mind: they just live and sense. They mostly rely on efficient chemical processes guided by a fine-tuned but hidden competence about homeostasis to “live a good life”. Sensing can occur without mind — the deck cites quorum sensing as the example [Diggle et al., 2007; Nealson and Hastings, 2006], where bacterial populations coordinate collective behaviour on population density.
Feeling. Multi-cellular, multi-system organisms exploit nervous systems to be aware of their internal state and of the environment around them. Nervous systems provide them with feelings, allowing creatures to represent in their minds the state of their own bodies — which is a required premise to homeostasis-driven behaviour. Feelings provide organisms with experiences of their own life: once anchored in the body frame, and oriented by the perspective provided by sensory channels such as vision and hearing, feelings contribute to the creation of a self [Damasio, 2010], a mental process animated by the state of the organism.
Knowing. Memorising experiences and anchoring them to the self leads to knowing and consciousness.
Now the central distinction of the chapter, stated as a comparison [Damasio, 2021].
Human intelligence is explicit. It requires a mind, as well as “mind-related” feeling and consciousness — which in turn require perception, memory and reasoning. Minds represent objects and actions through spatially mapped patterns, corresponding to our inside and outside; and the contents of the mind are inspectable and manipulable by the owner. Those patterns can be called images, with no direct reference to the visual sensory system.
Bacteria intelligence is hidden, non-explicit. Simply stated, they have no idea of what they are doing. They cannot construct patterns, they cannot represent objects or actions, and there is nothing there resembling reasoning. They have no mind — yet their intelligent behaviour works beautifully, based on well-articulated bioelectrical computations at the molecular level and below.
| Bacteria intelligence | Human intelligence |
|---|---|
| covert | overt |
| hidden, concealed | manifest |
| non-explicit | explicit |
| based on chemical / bioelectrical processes in organelles and cell membranes | based on spatially mapped neural patterns which “represent and resemble” objects and actions; imagetic |
And the clause that stops this from being a hierarchy: we actually benefit from both sorts of intelligence. Plenty of processes in the human body exploit “bacterial intelligence” — gut microbiota in digestion, metabolism, immunity. Humans exhibit every type of intelligence at the different levels of their inner operation.
Covert, hidden, concealed. Based on chemical and bioelectrical processes in organelles and cell membranes. No patterns, no representation of objects or actions, nothing resembling reasoning, no mind — and yet the behaviour works beautifully. It is homeostasis as a collection of how-to rules relentlessly executed according to no explicit directions, keeping the life-depending parameters within survival ranges. In Chapter 3 vocabulary: a goal that is pursued but not represented, that is the weak notion of agency.
Overt, manifest. It requires a mind, and mind-related feeling and consciousness, which in turn require perception, memory and reasoning. Objects and actions are represented through spatially mapped patterns — images — corresponding to our inside and outside, and the contents of the mind are inspectable and manipulable by the owner. In Chapter 3 vocabulary: explicitly represented goals driving the selection of actions, that is the strong notion of agency.
The two are not rival species but layers of the same organism. Plenty of processes in the human body exploit bacterial intelligence — the deck names gut microbiota in digestion, metabolism and immunity — and humans exhibit every type of intelligence at the different levels of their inner operation. For an engineer this is the licence to build hybrid systems without apology: a subsymbolic component that cannot explain itself is not a defective reasoner, it is a different and older kind of intelligence.
Do not read the bacterial case as a metaphor or a rhetorical flourish. The deck asserts something factual and strong: the behaviour is intelligent, it works beautifully, and it is grounded in well-articulated bioelectrical computations — while simultaneously involving no patterns, no representation and nothing resembling reasoning. If you soften it into “a sort of intelligence, in a manner of speaking”, the whole point evaporates, and with it the licence to count non-symbolic AI as AI at all.
The deck now supplies precise definitions, and they repay careful reading because ordinary usage runs them together.
Feelings emerged as a mental counterpart to a physical organism: everything we feel corresponds to states of our interior. They owe their existence to the fact that the nervous system has direct contact with our insides, and vice versa — interoception, which the deck distinguishes from proprioception (perception of our musculoskeletal system) and exteroception (perception of the outside world).
Feelings are interactive perceptions, built by the nervous systems — both peripheral and central — and by the cerebral cortex altogether. There is no direction: not from peripheral nerve components up to the cortex, with emotions possibly affecting the whole interactive process at any time.
The difference between external and internal perception is structural, not one of degree. Sensory processes such as vision and hearing lead us to mental representations of objects around us, a map of the world: the representation built by exteroception is a mapping, clearly separate from the world external to the organism. The interoceptive system works differently, because of the physical vicinity of feelings and the internal things felt, and also for the lack of physical barriers — myelin insulation, blood–brain barrier — allowing signals from the body to interact with neural signals directly during interoception. In short: feelings are both mind and body in the most extensive way possible.
No matter what the precise contents of a human mind may be — the landscapes, the furniture, the sounds, the ideas — those contents are necessarily experienced together with affect. What we perceive or remember, what we try to figure out by reasoning, what we invent or wish to communicate, the actions we undertake, the things we learn and recall: all of these processes can generate affective responses as they unfold. We can think of affect as the universe of our ideas transmuted into feeling — and, the deck suggests, it is helpful to think of feelings in musical terms: there, feelings perform the equivalent of a musical score that accompanies our thoughts and actions.
| Kind | Source | Examples |
|---|---|---|
| Homeostatic feelings | Constantly produced by the homeostatic processes and their interaction with the nervous system | well-being, hunger for food, thirst, pain |
| Emotional feelings | The collection of emotive reactions, weak or strong, that mental contents frequently prompt; part of the multimedia production that constitutes internal narratives | fears, joys and irritations of everyday life |
Both are essential devices in the generation of the conscious process, with homeostatic feelings working as the foundational grounding of our conscious being.
Consciousness is a particular state of mind resulting from a biological process toward which multiple mental events make a contribution.
Mechanically: the operations of the body’s interior, signalled via the interoceptive nervous system, contribute the feeling component, while other operations within the central nervous system contribute imagery of the world around the organism as well as of its musculoskeletal frame. These contributions converge, in a regulated way, to produce the encompassing mental experience of a living organism apprehending both worlds — within and around itself. The conscious process takes life within an organism, as expressed in mental terms, and locates it within its own physical boundaries; mind and body are given joint property of this ensemble.
And then the formulation the deck clearly prefers: consciousness as a gathering of knowledge. It is a gathering of knowledge sufficient to automatically generate, in the flowing of mental images, the notion that the images are mine — they are happening in my living organism, and the mind is mine too. The secret of consciousness is gathering knowledge and exhibiting that knowledge as a certificate of identity for the mind. Although integration does have a role to play when consciousness is conferred upon large numbers of images, consciousness is more than a mere integration of mental elements.
Notice how the boundary reappears here, in a completely different register from Chapter 2. There, the boundary was a membrane that constituted the identity of a cell. Here, consciousness is described as locating the organism’s life within its own physical boundaries and issuing a certificate of identity for the mind. Same structural move — identity by demarcation — performed once chemically and once mentally.
The deck ends where an engineering course ought to: with the questions this biology hands back to us.
There is much more intelligence than just rational intelligence. Can we precisely understand, capture and build any sort of intelligence? In case we can, are we able to use them suitably in our artificial systems? And — the question a designer actually has to answer — what sorts of intelligence do we actually need?
What have we achieved until now? Which existing contributions from the AI field can we recognise and place in the overall “living intelligence” framework just depicted? That is precisely the survey Chapter 6 undertakes: seventy years of AI, read against a notion of intelligence that does not privilege reasoning.
What about artificial consciousness? Do we need it, desire it, fear it? Could we aim at that? The deck leaves the question open, and so does this page.
The examinable core of this chapter is small and precise: (1) sensing as the most elementary sort of cognition, with the problem/solution framing of survival; (2) homeostasis as a collection of how-to rules relentlessly executed with no explicit directions, and its double role for autonomy and intelligence; (3) the explicit vs non-explicit comparison table; (4) the fact that humans run both. If you can also place emotions, feelings, mind and consciousness in their right relation, you can answer almost anything asked from this deck.
Because human concerns about intelligence started there, but that does not make it the right starting point for AI. The open questions are whether human intelligence is the only legitimate reference for designing artificial intelligent systems, and whether there is one type of intelligence or many. Going to living systems shows intelligence emerging as a critical property of life in the same way autonomy does — four billion years of intelligent behaviour, most of it without anything resembling reasoning.
The first prokaryotic single-cell organisms had no feelings, no thought, no reasoning, no mind and no consciousness, yet they sensed others like them and sensed their environments — where sensing is not perceiving, not creating a representation, not producing an image for the mind. They nonetheless responded intelligently to what they sensed, because the response helped their survival: the problem was potential death and the behavioural response was the intelligent solution.
An implicit, concealed competence accounting for the goal of maintaining life, with no logic, no reasoning and no representation. It is in charge of curating life according to the rules and regulations of homeostasis — homeostasis working as a collection of how-to rules relentlessly executed autonomously according to no explicit directions, yet ensuring that life-depending parameters such as nutrients, temperature or pH stay within survival ranges.
Chapter 2 presents homeostatic variables as an obvious manifestation of autonomous behaviour, and explicitly flags them as also relevant to the biological notion of intelligence. Chapter 5 delivers on that: homeostasis is the substrate of the earliest intelligent behaviour on Earth. One and the same mechanism is the elementary form of both autonomy and intelligence.
Internal: maintaining my own life — do I have enough nutrients, energy, building materials to perpetuate it? External: saving my own life, avoiding sudden death — am I in danger, is there an external threat, and how can I counter it and survive? Life is inseparable from the apparent goal of maintaining itself, with homeostatically regulated variables paving the way since the earliest single-cell organisms.
For functional coordination. Multi-cellular, multi-system organisms created a new level of complexity, and differentiated systems — respiratory, digestive, immune — called for coordination, which nervous systems enabled, improving homeostatic regulation and maintaining life more securely within mindless organisms. Mental images, feelings, consciousness, creativity and cultures came later, as a “secondary” effect.
Being: bacteria live and sense, with no nervous system and no mind, relying on chemical processes guided by a hidden competence about homeostasis; sensing can occur without mind, as in quorum sensing. Feeling: nervous systems make organisms aware of their internal state, providing feelings that let creatures represent the state of their own bodies — a required premise to homeostasis-driven behaviour — and, once anchored in the body frame and oriented by a perspective, contributing to the creation of a self. Knowing: memorising experiences and anchoring them to the self leads to knowing and consciousness.
Bacterial intelligence is covert, hidden, concealed, non-explicit, based on chemical and bioelectrical processes in organelles and cell membranes; there are no patterns, no representation of objects or actions, nothing resembling reasoning, and no mind — yet the behaviour works beautifully. Human intelligence is overt, manifest, explicit, based on spatially mapped neural patterns that represent and resemble objects and actions, and is imagetic; it requires a mind plus mind-related feeling and consciousness, which in turn require perception, memory and reasoning, and its contents are inspectable and manipulable by the owner.
No. We actually benefit from both sorts. Plenty of processes in the human body exploit “bacterial intelligence” — gut microbiota in digestion, metabolism and immunity are the examples given — and humans exhibit every type of intelligence at the different levels of their inner operation. The two are layers of one organism, not competing species.
Emotions are collections of co-occurring and involuntary internal actions — smooth muscle contractions, changes in heart rate, breathing, hormonal secretions, facial expressions, posture — triggered by perceptual events; they are usually aimed at supporting homeostasis, countering threats with fear or anger, or signalling successful states with joy; recalling events from memory also produces them. Feelings are the mental experiences that follow and accompany varied states of organism homeostasis, whether primary (homeostatic feelings such as hunger, thirst, pain, pleasure) or provoked by emotions (emotional feelings such as fear, anger, joy).
Interoception is the direct contact between the nervous system and our insides; it is distinct from proprioception (perception of the musculoskeletal system) and exteroception (perception of the outside world). Exteroception produces a mapping, clearly separate from the world outside the organism. Interoception does not, because of the physical vicinity of feelings and the internal things felt, and because of the lack of physical barriers — myelin insulation, the blood–brain barrier — which allows body signals to interact with neural signals directly. Hence feelings are both mind and body in the most extensive way possible.
First: a particular state of mind resulting from a biological process toward which multiple mental events make a contribution — interoceptive signals contributing the feeling component, other central-nervous-system operations contributing imagery of the world and of the musculoskeletal frame, converging in a regulated way to produce the experience of apprehending both worlds, within and around. Second, and preferred: a gathering of knowledge sufficient to automatically generate, in the flow of mental images, the notion that the images are mine. The secret is gathering knowledge and exhibiting it as a certificate of identity for the mind; consciousness is more than a mere integration of mental elements.