Rational Mind w Pietro Boselli Ep 5 - EMERGENCE
- The way we view and understand the world is strongly influenced by our language. Emergence is a term we don't often hear in everyday situations. And yet, the concept of emergence holds the key to interpret so many things around us and answer many different questions. If you listen to me long enough, you can recognize my voice and you have your own voice.
But what are voices? Are they a real thing? I have one and you have one so voices really exist, right? Or do they only exist as creations in our mind, after our brain converts vibration in the air into something we call a voice. Similarly, you can recognize color and describe things with the concept of color, but what is color? Are yellow and magenta real or just useful representation of a certain wavelength of the light? Similarly, if everything is made of atoms, is solidity real or just something we perceive? If neuroscience tells us that everything in our brain is the result of chemical interactions, then is free will real? And what is consciousness? Are the patterns we pick up in nature real or just useful fiction? Are all sciences equally fundamental or is a particular science more elementary than another? Intrigued to find the answers? The concept of emergence will shed light, not only on this but many more questions. (upbeat music) What is emergence? In its simplest definition, emergence is the manifestation of the properties of a system which do not belong to its individual constituents. Emergent properties are the result of these individual constituents operating together and at scale.
Emergence applies to many fields of knowledge, but let's focus on science for now beginning with an example. Consider a box containing a gas such as oxygen, air, helium, or anything else. You may remember from the school that a gas has measurable properties, such as pressure, temperature, and density. You might also remember that a gas is simply a collection of particles all moving around. If our example box was filled with the hydrogen gas, then there will be hydrogen atoms bouncing on each other and against the walls of the box.
Each particle has its own mass and at a given time also it has a velocity and a position. Now, depending on how the particle exist and interact the gas will have different properties. For example, density of the gas depends on how many particles are in a certain volume of the gas.
The temperature of the gas depends on their kinetic energy. The pressure of the gas is simply the force with which the particles would hit surface such as the walls of the box. So let's really think about that.
The individual particles themselves, for example, the hydrogen atoms do not have the temperature, nor pressure, nor density, but the gas which is made of these particles, does. The properties of the gas emerge from the interactions of the particles. They are emergent properties of the gas. The first question that arises is, are these properties real? Are pressure, temperature, and density of the gas as real as the velocity and mass of the particles? If I stick a thermometer in the gas box, I can measure its temperature. If I inflate the tire of my bicycle, I can definitely feel the pressure of the air inside.
But if I take a particle of the gas, there is no temperature, no pressure to be measured. So it definitely seems like temperature and pressure are real on a microscopic level even if they aren't at the microscopic level. Let's explore this further with a thought experiment, by introducing Laplace's demon.
In 1814, French mathematician, Pierre-Simon Laplace posited the idea of an all-knowing demon that could know the precise location and velocity of every atom in the universe. Such a creature would be able to exactly predict the future using the laws of physics. Laplace's demon is now used in philosophy for any thought experiment that requires being able to hypothetically calculate anything according to the known laws of physics. Imagine if I were an infinitely intelligent and all-knowing being such as Laplace's demon, I could know all the different velocities of each of the trillions of particles in the box, calculate the average kinetic energy at any instant and tell you the temperature of the gas. Or I could just be a random dude who knows nothing of the particle, sticks a thermometer in the box and also tells you the temperature. This is the first special thing we discover about the emergence.
It lets you ignore a lot of information, and still gives you accurate predictions or representations of the world but only for particular cases. If I know the temperature of the gas, I can ignore the kinetic energy of the trillions of individual particles. But if I don't know the temperature and I know the kinetic energy of all the trillions of particles, except for one, I cannot tell you anything about the temperature even if I am as intelligent as Laplace's demon. How is this useful or relevant? Well, suppose I am only interested in how the temperature and pressure of the gas change as I heat it up, I can do this very simply with some elementary formula from thermodynamics, without having to somehow know the velocity and position of each particle in the gas. So if you are only concerned with thermodynamic properties such as temperature or pressure, it doesn't matter that particle physics can explain how these properties emerge in the gas. There are much simpler ways to measure these properties.
Thermodynamics is therefore a science that emerges from particle physics. But is it just as fundamental? Can we learn something about the world that we wouldn't by knowing just particle physics? Let's take a step back. Consider a science such as cognitive psychology, which can describe things such as heuristics and biases, our behaviors and so on. If we wanted to go deeper and understand how these human behaviors emerged, one would look to neuroscience.
There we will learn how neurons and neurotransmitters and receptors interact and affect our behavior. But the working of neurons could be further reduced to cell biology, which in turn is reduced to molecular biology and so on through chemistry, then physics, then elementary particle physics. It is almost like there is a hierarchy of sciences.
This is known as the reductionist hypothesis, everything can be reduced to a more fundamental explanation all the way down to elementary particle physics. But what is the implication of this? Are the elementary particle physicist the only ones studying anything fundamental and real? The answer is clearly no. For example, biology has no interest in particle physics and yet biology is a science in its own respect that can explain many real phenomenon and biology results in real technologies such as medicine.
It would be impractical if not impossible to try and explain how a drug works by calculating how all the single atoms interact in each individual molecule of each individual cell for all the tissues and organs that make up an organism. This is another thing we discover about emergence. Even if there is only one reality, there are multiple ways to describe it, all equally accurate, at different levels.
And this is true not only for the way we scientifically describe the world it is also true when it comes to our perception of the world. For instance, think of my example from earlier about voices, the property we call a voice emerges from vibrations in the air interacting with our nervous system and drawing on our learned conceptions and memory. We also see an emergence when we consider our nature has evolved to take advantage of the higher level patterns that emerge from lower level ones.
Swarm behavior is probably the most cited example of emergence nature, and it is apparent in the flocking of birds or in schooling of fish. Flocking is considered an emergence behavior as there is no master coordinator instructing each bird what to do in order to fit in the grand scheme of the flock. Each individual bird is only aware of its immediate surroundings and follows three simple rules, separation, alignment, and cohesion.
These rules mean the birds keep the correct distance and alignment with their immediate neighbors. The resulting flock movement and shape emerged from all these interactions but there is no awareness of the whole on the part of the individual. Another widely cited example of emergent behaviors comes from ant colonies. Individuals ants are not intelligent enough to survive and in fact, even a small group of ants removed from the colony will just run around until they will starve.
And yet each ant follows some simple rules relative to the peers in the colony. Thus the coordinated action of the colony in gathering food and building a nest emerges from the individual actions without central control or a mastermind coordinating at the top. Human society itself shows emergent behaviors.
Despite the fact that humans are the most intelligent animal on earth think about how helpless an individual human would be if isolated from society. Even the most educated and intelligent of people couldn't by themselves create the food we have, the houses we live in the technology we enjoy and the medicine that ensures our health. And yet collectively we are capable of incredible things like sending spacecraft to other planets. So as humans we organize in societies and new emergent behaviors and actions arise, which do not exist at the individual level. Describing the world by a higher level observations that emerge from deeper and more intricate ones, and living out information from other levels does not mean we are discarding the richness of reality. This is because at each level there are patterns or properties that are not observable at another level.
Think about the concept of solidity. If I have a table in front of me I can certainly touch it and describe its properties such as roughness, smoothness, hardness, or color, but the molecules making up the table do not possess any of these properties. And similarly, molecules of water do not possess properties of wetness, but water itself does. In fact, the same molecules when they undergo a phase change like from water to ice display completely different emergent properties. They transition from wetness to solidity with exactly the same individual constituents, but just slightly different interactions between them. This is just to say that by observing the microscopic, it is impossible to predict the qualities that the macroscopic would have.
But by observing macroscopic behavior it is possible to see how the microscopic is causing that behavior to emerge. Basically the reductionist hypothesis does not work in reverse, nothing in the water molecules speaks about wetness or solidity because these molecules do not possess these qualities, only water or ice do. Given that none of us is Laplace's demon, emergent properties allows us to talk about reality in a useful way, and also to make accurate predictions. Nature has evolved organisms, including humans to be a ruthless pattern pickers.
We have certain heuristics which are subconscious mental shortcuts that allow us to make sense of the world without having to know the microscopic interactions at play. Let me give you an example. Suppose I throw a baseball towards a window, Laplace's demon could predict the outcome of this by instantly calculating exactly the neuro impulse controlling my action, the chemical reactions contracting my muscle fibers, the resulting arm movement, the force on the baseball, it's acceleration, the air resistance, the exact trajectory, the momentum against the window, the harness of the window based on the bonds between molecules, the propagation of the cracks, the window falling apart and succumbing to gravity. By knowing all the information about all the atoms it could make all these calculations all the way down to the quantum mechanics of this event. And yet, the prediction of a little boy observing the event would be the same, the baseball is going to smash the window.
The boy did no calculations and in fact, he doesn't even know how he came to that conclusion. His brain is the result of an evolutionary system that took advantage of emergent patterns. The ability to pick up pattern is known as our intuition also known as common sense. This ability is something that humans and animals share, and that machines struggle to reproduce.
Our ability to reason, conversely, is the little Laplace's demon discovering the low level structure and working of things. Our intuition is fast, our reason is low, and this is why we heavily rely on our common sense, less on our reason. And sometimes we see patterns where they do not exist. This is especially true of intentionality.
Intentionality is the state of mind reflecting a deliberate action directed towards a specific end. There is a type of gazelle in the savanna that when chased by a lion start jumping around in big fancy and energy consuming leaps. This seems strange. Why would it do that instead of just running? The lion it turns out, chooses to chase the other gazelles in the herd, the ones that are not jumping around like crazy. It seems as if the gazelle is showing off to the lion, "Don't bother chasing me because I'm so fast that I can afford to show off, so you have no chance if I actually started running flat out."
The lion seems to pick up this message and diverts the attention to the weaker gazelles who have no time to show off. Now, we humans, seem to immediately ascribe this intentionality to the gazelle, it is showing off to give a message to the lion, but does the gazelle know that that is what it's doing? It does not. The gazelle is simply the beneficiary of a rational system that we humans, can comprehend, but the gazelle does not. What the gazelle displays is just emergent behavior resulting from evolution. It's instinct automatically commands big fancy leaps at the sight of a lion, and the lion as a result chases another gazelle. The intentionality of the gazelle does not exist as far as it's concerned, but it does as far as we are.
This brings us close to another deep philosophical implication of emergence, the nature of consciousness. Is there something special about conscious beings? Let's begin exploring this by considering intelligence. A person can be set to possess intelligence, the ability to process information and understand concepts. Intelligence can be seen as an immersion property of the interactions of all the neurons in the cortex of the brain. The way the individual neurons interact allow the understanding of a concept for example, the concept of a tree or the concept of brightness. Even if the individual neurons do not understand this concept.
So intelligence is emergent because individual neurons do not possess it in the way the person does. Intelligence arises from the interactions of neurons acting as logic gates for electrochemical signals in the human nervous system. But what if the signal was gated through silicon semiconductors instead of brain cells? In a traditional computer program, everything is coded permanently by a programmer, but with Artificial Intelligence, the parameters are adjusted by the AI program itself. Learning data is fed to the AI which compares the output to the desired behavior and then adjust the parameters again and again until an intelligent task emerges. On a theoretical level, intelligence could emerge from lower level logical operations independently of whether these are operated on biological neurons or on silicon. This is known as substrate independence.
Substrate independence applies also to other systems. For example, one could have two glasses with the same emergent microscopic properties, but different molecules. So one could have the same intelligence emerging from either biological neurons or artificial neurons, whatever this may be. The first objection that arises when talking about intelligent machines is the one about consciousness.
Somehow it is often believed there must be a special quality to consciousness, that it couldn't simply emerge from the chemical interactions at the neuron cellular level, nor at the molecular level, nor at the atomic level, but why? A useful thought experiment is imagining the philosophical zombie. A philosophical zombie is a human that instead of being born is hypothetically build, bit by bit by adding all the different molecules to form all the different proteins, cells, tissues, and organs. The philosophical zombie could work in all its physiological functions exactly like a human, would this being be conscious? Some, who attribute a special quality to consciousness say the answer to this question is no, thinking you either have consciousness or you don't. And yet if you ask the philosophical zombie, do you have consciousness? It would say yes, given that it lives, breathes and thinks exactly like a human. So where is the difference? How do we know we are not ourselves philosophical zombies? There seems to be no difference, and so we draw the conclusion that consciousness is not a special quality hidden somewhere but itself an emergent property of the organism in question.
That means that consciousness is not on or off, but more of a continuous property. Is a human conscious? Yes, of course. A dog? Yes, a dog is very conscious, maybe not to the same degree as a human, but still. Trees interact with their environment and are capable of adapting and changing according to external stimuli. So it could be argued that trees possess some lower degree of consciousness too. On the micro-level, are neurons in the brain conscious? To an extent, yes.
The brain cells interact and communicate and adapt to their environment. So consciousness is not just on or off, it is not a special property, but it emerges from lower level interactions. The temptation to attribute a special status to things that can not be precisely explained by current science has been around since the beginning of human knowledge. Even the most intelligent people who ever existed have done so at some point. As scientific understanding advanced, it was often discovered that those special properties were not so special after all.
Democritus, who theorised the existence of atoms as early as 460 BC, believed that only the atoms, the indivisible component of all matter, were real. Everything else was just by convention, for example, color, hot and cold, sweet and bitter. This is because there was no scientific explanation of how color or sweetness arise in the experience.
So this qualities must have been something other than strictly speaking, real. Galileo, the father of modern science believed similarly, the things like odors and colors and tastes only resided in the subjective experience. Descartes, the founding thinker of rationalism also believed in this duality. So did Isaac Newton himself, who invented calculus and managed to syntactically explain so much and demystify so many concepts.
Newton believed that some qualities were primary and could be explained scientifically and that some were secondary. Newton believed that things like mass were primary qualities, but things like color was secondary. But then James Maxwell came along with the equations of electromagnetism. Then it was understood that light could be described as an electromagnetic wave or radiation and that different frequencies of this radiation result in different colors. So once again, something considered to be other, in this case color, was revealed to be just an emergent property of a more microscopic interaction.
A similar approach can be applied to free will. The question is, if our behaviors and actions are just the deterministic outcome of our physiology and our thoughts, beliefs, and even cautiousness emerge from the chemical and molecular processes happening in our brains, then where is the free will? One could argue there is no free will. Then should we not be held accountable for our actions? After all we're just driven by the forces of nature, and our actions are the inexorable result of the laws of physics setting motion at the inception of the universe. One could make this argument in front of a judge and say, "It is not my fault your honor, free will doesn't exist." But this argument wouldn't hold because as we have seen, even if something can be reduced to its more fundamental components, it doesn't make it any less real. Just like the gas in the box is a gas with a temperature and not just a collection of particles.
Just like intelligence exists even if it emerges from neurons. Just like an ant colony or a nation composed of individuals are real so is free will. It could be considered a useful fiction, but as far as our understanding of the world goes, either everything is a useful fiction or everything isn't. If electrons are real then so are we. So grasping the concept of emergence really helps us better understand our reality and answer many profound questions, but does it have some more practical applications too? In a way we are already using emergence in computation.
For instance, in algorithms such as artificial neural networks, the learning emerges from parameters which are self adjusted by the learning algorithm. So it is fair to say that the macro behavior of the AI emerges from the lower level interactions. Similarly, we can arrive at computational solutions to problems via things such as genetic algorithms.
These algorithms evolve parameters to get a desired outcome by applying rules inspired from biology and evolution. All these computations are still based on a single algorithm orchestrating all the individual low-level operations which run on the processing unit. So we are setting simple low-level rules and we get the high level results, but there is no explicit rule for mapping from one level to the other.
In other words, there is no full understanding on how to directly manipulate the low-level rules in order to achieve a desired outcome. We have models to simulate neural activity and biology, which worked well and give us a result, but we do not fully understand how the models work. Here to explain further is Professor Dana Randall, the Co-Executive Director of the Institute for Data Engineering and Science at the Georgia Institute of Technology, with a special focus on randomized algorithms. - I'm asked this all the time. Why is it important to prove things if you know it's going to behave like this physical system, why can't you just use it? And many people do, and I have no problem with that work, but by understanding it and really understanding the very simple rules and how they then lead to this emergence, you understand what flexibility you have. You can really have a lot more robustness and flexibility in your model when you understand exactly why liquid is turning to solid.
- Could there be a way to model emergence and rethink computational together? Where to start? First thing to do is to understand simple rules of emergence, recreating simple systems which can be inspired by emergence nature. - Many scientists have made great advances in trying to understand what the ants are doing, how they're communicating, kind of how their programed? But you can take a different approach and not try to understand what the ants are doing, but you could ask how dumb could the ants be and still accomplish what they're doing. Can we come up with a very simple model of computation that mimics the behavior of ants that is a prove that they don't even need some of the intelligence that they have? And we could accomplish and mimic kind of interesting group behavior with very, very simple rules. - Our first step is to create swarms of dumb robots capable of performing only very simple low-level interactions between each other and see how this results in an emergent collective behavior. - So if you wanna get swarms flocking, each one is seeing just its immediate neighbor and they have a preferred distance so you don't want them too close, you don't want them too far and you want them roughly to align in the same direction.
So this is a model that we're studying mathematically now so you have each one can move and spin in different directions but they're more likely to wanna align with any neighbor. - But this approach can be taken a step further in order to try and get the desired behavior and perhaps discover a new way of harnessing emergence to get a specific outcome. - We're trying to understand a mathematical model of collective intelligence of how you can have individual particles that interact, but only see their immediate neighborhood and come together and collectively are performing some tasks.
So you can do that with a puppeteer who is choreographing and telling everybody where to go. But when you look at ants or flocks or lots of biological systems, there's no master puppeteer. So we're trying to understand how we can get very simple programs that caused these things to interact, have very little memory, little communication, they don't even know where they are, which way is north' they know very little. And we're trying to see what we can prove that we could get out of these systems with very simple rules. And the reason we're doing that is that we're hoping that we can come up with what we're calling algorithmic matter.
- Algorithmic matter could open the door to many new technological applications. - So you have these tiny cells that can now be programmed with very basic collective behavior. Well, if they're the size of a red blood cell, we know that they could be injected into the bloodstream. So we have the potential of building more and more sophisticated algorithms that are still quite simple, but can go to a place where there's some anomaly, some injury and potentially do some healing, some self healing of a system and repairing.
- And beyond this being able to control how collective intelligence emerges from simple low-level interactions can potentially lead to reinventing computational methods. - I think that a lot of people are beginning to look at collective intelligence and emergence because computers are reaching their physical limitations. Moore's law says that computers will be roughly twice as powerful every year, and this was sustained for decades. But now we are at what's called the thermodynamic limit of computation.
We know that physical constraints say, you cannot keep doing same old and getting more powerful. So you really need a very different approach. So some people are looking at quantum algorithms, some people are looking at self assembly as a different kind of emergence.
And there's some sense that we could possibly think about computation very differently. - If you'd like to learn more about any of these issues, check out the link below or visit www.rationalmind.show for more information and resources. (upbeat music)