Intelligence is not as rare as we like to believe
Once life starts I think intelligent life is a very likely outcome. It has evolved separately several times on Earth.John Hardy, during the 2019 Breakthrough Prize panel discussion, “Is there life elsewhere in the universe?“
It is posited (dare I say accepted?) that intelligence has evolved independently more than once on Earth. Once (or three times) for marine mammals, elephants, and primates. Once for the crow family of birds and parrots. And once for cephalopods, specifically the shell-less Coleoidea: octopuses, squid, and cuttlefish.
For me, the idea that non-human animals are intelligent is news. I suspect this blind spot grew out of a desire to avoid emotional stress, as I have been eating animals for a lifetime, and sacrificed many thousands of animals for neurobiology research. It is hard to do those things while recognizing that animals are intelligent. I am also a person focused on words and communication, and I tend to shut out beings I can’t communicate with, human or not.
My changing point of view all started with a chance glimpse of a book spine at the library. The title was, “Are We Smart Enough to Know How Smart Animals Are?” by Frans de Waal. I laughed. I didn’t pick it up or read it, but the title stuck with me, and I started to consider the question.
Once I started looking for it, evidence of animal intelligence was everywhere. See, for example, this excellent article by Jonathan Balcombe in Nautilus, “Fish Can Be Smarter than Primates.” Watch any episode of Planet Earth or Blue Planet, and you will be stunned by the number of species that use tools or join forces with others to coordinate effort toward a common goal.
Orcas working together to stun and eat fish, via GIPHY
How is animal intelligence defined and measured, what evolutionary pressures are believed to result in intelligence, and what, if anything, makes human beings special? Below is what I have found out so far.
What is intelligence?
We can all list features of human intelligence: Learning, planning, goal-directed behaviors, decision-making, problem-solving. More generally, we call this executive function. We believe our cognitive powers arise from the cerebral cortex, the outermost layer of the brain.
Looking at the brain macroscopically, we notice it’s convoluted surface. Gyri and sulci create crests and valleys; more surface area for more neurons.
Microscopically, we see that the cortex is organized into rows and columns.The rows are like a six-layered cake, each layer with its own ingredients and flavors. Cortical columns are individual computational units, each one dedicated to processing a specific sensory input, such as the upper right corner of your visual field, or the sensation of your pinky finger. The columns are not isolated. They connect to each other and to deeper brain structures in the brain.
This is all great information. We can list some behaviors we think of as displays of intelligence. We have a parts list and brain map that are getting annotated in greater detail every day. We are even starting to link the two, identifying the neural circuits underlying perception, decision-making, and goal-directed behavior.
All of this information still fails to answer the question, what is intelligence, especially in a way that is generalizable to non-humans. That is precisely what we need to figure out before we can recognize intelligence in other beings, who will not have the same brain structure or behaviors as us.
It appears that both animal researchers and AI developers struggle to define what counts as intelligence. George M. Church, Professor of Genetics at Harvard, just posted an excerpt from his chapter of “Possible Minds: Twenty-Five Ways of Looking at AI” that I think is relevant here. He admonishes that humans are hard pressed to acknowledge artificial intelligence as valid or worthy of protection, and they should. Sounds a lot like discussions about animal intelligence.
He provides an insight that may help us define intelligence. He is talking about algorithms designed for AI systems here, but I think it applies more broadly to the evolution of intelligence.
For free will, we have algorithms that are neither fully deterministic nor random but aimed at nearly optimal probabilistic decision-making. One could argue that this is a practical Darwinian consequence of game theory. For many (not all) games/problems, if we’re totally predictable or totally random, then we tend to lose.George M. Church. View entire excerpt, “A Bill of Rights for the Age of Artificial Intelligence: We should be concerned about the rights of all sentients as an unprecedented diversity of minds emerges.”
I like his phrasing here that free will, or probabilistic decision-making, is a Darwinian consequence. Unpacking that a bit, there is an optimal balance between predictability and randomness that enables beings to be flexible in thought and action. It is required for innovation. Having the capacity to try something new in challenging circumstances can be the difference between survival and extinction.
I think that optimal probabilistic functioning is the core mechanism required for intelligence. It is a feature at every level of organization, from behavior, to the neural circuits underlying behavior, to the proteins underlying neuronal signaling.
Electrical signaling in the brain is based on the opening and closing of an array of ion channels embedded in the membrane of neurons. Guess what? That opening and closing is stochastic. There is a probability distribution that an ion channel will be open given a certain stimulus. Even for ion channels where the structure has been determined at sub-nanometer resolution and the physiological mechanisms that cause the channels to open and close have been figured out, no one can predict with certainty whether or not an individual ion channel will be open at a given time.
Optimal probabilistic functioning is another way of saying that the system functions with an ideal probability distribution. The tails of the distribution are in a sweet spot where there is just enough randomness, just enough noise,so that the system can routinely try out alternative signaling patterns in the background without disrupting the functioning of the whole. Sometimes randomly generated new signaling patterns produce results, and are positively reinforced. For evidence of this, read about how the brain learns to control external devices through brain-machine interfaces.
Do I sound crazy, or does that make as much sense as I think it does? Ok, let’s ground that big reach for a definition with a more practical question — how do animal researchers measure intelligence?
Indicators of animal intelligence
There are several characteristic features of brain structure associated with intelligence. One feature is a large brain relative to body size. The second is the presence of specialized structures associated with executive function, such as attention, planning, and learning. As mentioned, human beings have the cortex. Birds have the nidopallium and coleoid cephalopods have the vertical lobe. Another is a high density of interneurons, essential for creating both local and long-range connections between specialized ganglia within the brain and peripheral nervous system.
These accepted indicators are biased by our belief that human beings are the pinnacle of intelligence on Earth. I am inclined to think the nervous system is flexible enough to configure itself many different ways to give rise to intelligence, and we should not close our minds to alternatives.
Cephalopods are a good example of that. Though they have all the indicators listed above, they also have unique brain features. Cephalopods control the movements of many flexible limbs with a huge number of degrees of freedom. To do so they rely much more on processing in peripheral neurons so that they can execute stereotyped movements without relay to the central nervous system. In fact, they do not appear to have any central representation of their limbs like intelligent vertebrates do (see this article about Octopus locomotion).
The nervous system of cephalopods represents a striking example of embodied organization, in which the central brain acts as a decision-making unit that integrates multimodal sensory information and coordinates the motor commands executed by the periphery.From Piero Amodio et al
This octopus travels with a coconut shell for protection from predators and to lie in wait for prey, via GIPHY
In addition to brain morphology, scientists observe and test animal behavior to look for behavioral flexibility, or an animal’s ability to change their behavior based on circumstances. Some examples include demonstration of learning, problem solving, planning, or tool use, especially innovative tool use or the simultaneous use of multiple tools. Social animals may demonstrate their intelligence by working together to achieve their collective objective. There is also a link between engaging in play and intelligence, though there is debate over what comes first (do intelligent animals play, or does play make animals more intelligent?).
In contrast, here are some behaviors that are not considered intelligent: Stereotyped, repetitive behaviors. “Hard-wired” or trained stimulus-response behaviors. Trial-and-error to achieve a goal, rather than problem solving.
There is some debate about whether or not behavioral flexibility is enough to label a species as intelligent, as such behaviors can be supported by simple neural circuits. But if the intelligent behavior is there, does it matter that the neural circuit supporting it is simple? Intelligent is as intelligent does.
Intelligence evolves when it becomes necessary for survival
Brain size be damned, if it’s critical to a species’ survival then that species will most likely be good at it.Jonathan Balcombe, “Fish Can Be Smarter than Primates”
Cuttlefish use a light display to stun prey, via GIPHY
In a recent Trends in Ecology article, “Grow Smart and Die Young: Why Did Cephalopods Evolve Intelligence?” Piero Amodio et al. describe three selective pressures that are believed to contribute to the evolution of intelligence.
- Challenges in finding and processing food (The Ecological Intelligence Hypothesis).
- Challenges of group living (The Social Intelligence Hypothesis).
- Challenges of predator-prey interactions.
Cephalopods are different from other intelligent species. They are invertebrates, have short life-spans, only mate once, don’t care for their young, and are not social animals. Yet they have the brain structures and behavioral flexibility of intelligent vertebrates. Why did they evolve intelligence?
Piero Amodio et al. argue that the loss of their shell 275 million years ago caused the Coleoidea group of cephalopods to evolve intelligence due to increased vulnerability to a wide range of predators.
Sudden vulnerability to predators can drive rapid evolution
A few weeks ago, an ambitious experiment demonstrating natural selection was published. See the research paper, “Linking a mutation to survival in wild mice,” by Rowan DH Barrett et al. in Science, and a great example of science journalism done right, “The Wild Experiment That Showed Evolution in Real Time,” by Ed Yong, in The Atlantic. I especially loved Ed Yong’s article because I have spent time in Valentine, Nebraska, where the research was completed, and he provides an amusing and uplifting view of interactions between the locals and the researchers that reflects my experience there.
For this experiment, researchers captured hundreds of wild mice with different colored fur and placed them in large enclosures that were either built on light-colored sand or dark-colored soil. After just three months, many of the mice with fur colors that didn’t blend with their environment had been eaten by owls.
The researchers sequenced the Agouti gene, known to contribute to fur color, for each mouse. They discovered seven Agouti mutations resulting in variations of coat color, and were able to correlate each mutation with the probability of survival. They found a mutation that made mice much less likely to survive on dark-colored soil. Unsurprisingly, the mutation resulted in lighter fur, making those mice easy for owls to target on a dark background. In a single generation, that mutation became common in the population on light sand and rare in the population on dark soil. They are continuing this experiment to see how this pans out for future generations, and to sequence the entire genome of each mouse to look for other genetic variations that affect survival.
Different types of changes can result in sudden vulnerability to predators. In the case of the light-colored mice, their environment changed. They had previously found a niche optimal for their survival, and were forced into a new niche that they were genetically unsuited for. This makes me think of the sad truth that many environmental niches are being destroyed by human activity. Species that have evolved to survive in those niches are suddenly displaced and many of them will die.
To generalize, sudden change in availability of niches (gain or loss), combined with an increased mortality rate due to predators, exposure to the elements, or infection, can cause rapid enrichment of certain genetic traits. This only works if there is genetic variation in the population and at least a small subset of the population has traits that help it survive. If not, the species will become extinct.
In the case of the unarmored cephalopods, their body shape changed. They had to utilize new niches previously unaccessible to them, and develop clever behaviors to evade capture. The massive die-off they must have faced likely resulted in a rapid enrichment of certain favorable genetic traits, including intelligence.
Are human beings special?
Yes, of course. We are arrogant in assuming that our intelligence far exceeds that of other animals, to the point that we have a hard time recognizing animal intelligence at all, but human beings clearly have an intellectual advantage. An interesting question is, what makes us different from other intelligent species?
I believe our ability to pass down information from one generation to the next is what makes us able to achieve so much more. The information we contribute to and curate in our lifetime lasts beyond our biological existence, such that the next generation may take what we have found and continue to build upon it.
In The Only Harmless Great Thing, author Brooke Bolander imagines the inner lives of intelligent, matriarchal elephants. She beautifully states the importance of shared wisdom from the perspective of an elephant:
Without stories there is no past, no future, no We. There is Death. There is Nothing, a night without moon or stars.Brooke Bolander
Indeed, stories are a way to pass information from one generation to the next, creating communal knowledge, or “We.” Human beings likely started to share information through oral tradition, passing important stories from one generation to the next through word-of-mouth.
Eventually we developed technologies to preserve stories and information outside of our minds. The earliest evidence of written records are inscribed stone tablets dated 3500 BCE. Next came papyrus, dated 2500 BCE. Jump to present day, where our information technology not only records information but augments our abilities to recall and process information far beyond anything our brain could do.
The intelligence of human beings is no longer constrained by biology, and our “We” is more than any one of us could learn in a lifetime. The night is bright with stars that we have mapped and studied through many generations, and that information is accessible to any who wish to learn it.
If a dolphin could hold a pen, the same might be true for them. Of course the precursor to stories is language. Animals clearly communicate. Is their language limited to warnings of predators, attraction of mates, and location of offspring, or do they have more to say?
It is believed only a handful of species are able to learn new vocalizations: humans, dolphins, whales, seals, elephants, bats, and several species of birds. This number may grow as we start to look for evidence of vocal learning in more species. Vocal learning is the foundation of language development. Only if we are able to decode animal language will we be able to know if they have stories to tell. If they do, perhaps we can hold the pen for them.