Consciousness is a simulated state; it can only exist in dreams, not in [Music] physics. What do we mean by consciousness? The confusion that we have with consciousness is that we think it's must be a physical thing because it's really

existing, but for to really exist, things don't need to be physically implemented. For instance, money is not physical, right? Yet you cannot explain our reality if you assume that money doesn't exist. It is implemented in reality; it's implemented using uh, little sheets of paper that we print numbers on, in bank

accounts, in computers, and so on and so on. And we use it to structure reality. It has causal power. Money is a pattern with causal power that exists because we project it in a stable way into the world, right? And money exists as if it was real. It's virtual in a way, but it's

implemented in such a way that it works. And our mental states are in the same category. They are virtual. There is exist as if, if you zoom in and you see the matter interacting, you see just these patterns in the activation between neurons. Neurons are not conscious; they're

physical mechanisms, right? Brains are not conscious, but for a brain, it would be very useful to know what it would be like if there was a person that cared and perceived everything as it happens. And so they create a simulation of this that exists virtually as if a dream, and

we exist in that dream. Consciousness is a simulated state; it can only exist in dreams, not in physics. And this is this big conundrum that we experience. We notice that we exist in a dream, right? It's a big conspiracy. Magic is possible, and yet the physicists tell us that it's

Not physics, it's talking about a parent universe that is made out of arcane mathematics, quantum mechanics, and so on. We can never visit this parent universe because in physics you cannot be conscious, you can only be conscious in a dream, and physics has the property of

creating brains and organisms that can dream, and machines of other types. Our perception of reality is a trance in a dream. We can deconstruct it. If you wake up, we realize it's just a representation. So, but can we compare our digital computers at all to human brains?

And at super different, and there are human brains. Not too complicated, like an individual neuron is so complicated that you would need at least a 12-layer neural network to understand what it's doing in with respect to its input and output, right? So, how many computers would

it take to emulate a human brain? If an individual neuron is already so complicated, this would be an astronomical number, right? But we also need to ask from the other direction: How many brains would it take to emulate a computer like this? How many human brains

would you need if you want to run Meco without flaws? Also an astronomical number, right? Because our brain is very mushy, very slow, very noisy, very unreliable, and so when we want to get a better handle of how to compare this, it usually takes an example like this: When

stable diffusion, uh, came out, the Bates of Stability AI, that was learning a diffusion model to make sense of all the visual data. They put millions of pictures into this and found the statistical regularities in them, so you can prompt this and generate an arbitrary picture. And this model

contains an entire visual universe at a much higher fidelity in depths and variety than a human brain. It has all celebrities, all dinosaurs, all spaceships, all artistic styles, everything is in there. You can download it. It's the first version was 2

gigabytes. It's quite depressing. 80% of what your brain is doing is in 2 gigabytes and it's actually much more. So uh, to just to get an idea of how simple we actually are, we're just very very efficient using this mushy

substrate. So of course these systems are designed in a very different way. We have an outside-in design and computers, they stabilize by our sub that we build deterministically. So it's behaving in a very well-fashioned way, we know what it's

going to do in every moment and uh, the training is decoupled. We give it static data, it's not interacting with the world while it's training. It is based on a prediction paradigm and it uses machine learning training algorithms. And our mind is very different. Our brains are

using an inside-out design. It's self-organizing, it's coupled with the environment in real time, it's optimizing for coherence and it's continuously developing itself. So tech, technological design is this outside-in where you have a desk where you understand everything, you understand all your tools and then

you are extending this by building an additional mechanism that extends your desk and gives it more functionality. And in nature, this is the other way around. You have a seed for instance, the seeding of a tree that is by itself not a tree yet, but it wants to become a tree and to

do this, it needs to colonize a chaotic environment, subdue it, turn it into something that it understands that it can work with. And one way of doing this is it splits into lots of units that behave pretty much the same, so they become predictable, and then it creates

some kind of scalable colony that is becoming larger and larger, and this is how this Inside Out design works in nature and also in our own mind. If you look at the individual neuron, these are mouse neurons, embryonal ones, you see they are all single cells

animals that try to survive, each of them self-directed, and they have to link up and find an organization among each other that works, or otherwise they will start to death. And this is how it works, it's completely inside out. And there so there was a branch in computer science

that looked at these principles of self-organization. So far it didn't get really far, but it already started with Turing, who looked at reaction diffusion patterns, chemical reactions that give you regular patterns that can be dynamic, and looked, can we use them to understand

computation and self-organizing systems? And uh, then some people like Alex Morin, who's working at Google in Zurich, build simulations that are more modern that uh try to generate such patterns in a general case, and um, then I come to this idea of cellular automata by Conway and

others where we build computation in a systematic way out of simple uh systems that only look at the state of their neighbors of their environment, and then change state accordingly to what they observe in their environment. And the biologist Mike Levin is working with

This expanding this into neural cellular automata, they explain development in organisms, but also possibly the self-organization of our brains. And so we can already take this to learn, for instance, textures and so on. It would be really interesting is to try if you can learn arbitrary functions to control behavior.

With that self-organizing principles and uh, the uh neuroscientist Jerry Adelman suggests that our genome is not encoding very specific architecture for our brain. Instead, in everybody's mind, there is an evolution going on, an original competition between different forms of governance. And there is only enough

information in the genome to bias that competition, so it converges relatively quickly into the right architecture. So it's every mind is an evolving system. Everyone has an individual evolution of our mind for structure. And I suspect that consciousness is very, very instrumental there because it's this main principle

of the organization. We typically think that consciousness is extremely sophisticated. Maybe only we have it. It's super rare in nature. Uh, it's really the pinnacle of things. But you don't get conscious after the PhD, right? You become conscious before you can track a finger.

So uh, you also noticed if somebody is not getting conscious as an infant, they're not getting anywhere. They remain a vegetable. And while we are not conscious, we cannot actually learn, right? So it seems to be a simple learning algorithm in nature for a self-organizing system. That's a hypothesis. I

don't know that's true, but I think it's very tempting to think like this, that consciousness comes first, not last in mental organization. And uh, so maybe it's more simple than perception and thinking and everything else. And all the other stuff comes from the organization that

Consciousness, uh, imposes on it. It's much more ubiquitous in nature than we think, and maybe uh we are not the first ones to discover this. And I suspect that this theory has always been hidden in plain sight. There is a text that is uh

the oldest known copy of it is um something like 100 to 200 years before Christ, and uh it was integrated into the Hebrew Books probably around 500 BC, but it might be uh a lot older than this apparently. And this uh text is the first

chapter of Genesis. And if you ever read this text, it's very cryptic, much more than the others, right? This describes this H Elohim hovering over this uh waters before there's anything else, where there's no no space, no time, no light, no darkness, but there's already

water. And then it builds a firmament that separates the waters above the firmament from below the firmament, and you know how the whole thing ends. It creates us in its own image. I don't like anything like a spirit hovering over the waters, right? Uh, what does this mean, uh?

also, where are these people confused, right? Telling a story about the creation of a physical universe by a supernatural being. What kind of experiments do they claim to have made? Did they talk to a burning bush? In which universe is this a valid experiment? But who would believe

that maybe this is there something else? Maybe the story actually means is a six-step theory about the emergence of uh consciousness and cognition in a mind, because the universe that we are in is a dream, right? We dream this universe that we are in. There are no colors and sounds

and emotions and facial expressions in physics, they all exist in our minds as ways to make sense of reality and the physical Universe. Back then, when this text was written, this was not a thing. People hadn't discovered this idea of describing the Universe with short differential equations. Instead, they

realized we are in a dream, and that so they observed in themselves and in others and their children how the stream is forming. And so this all starts out with Consciousness hovering over the substrate. We now know that the substrate are neurons. They didn't know that, so

they just used the word substrate. And the next people didn't know what substrate means, so they translated it as water. And so Consciousness forms on the substrate, and the substrate is analyzed into, and then we make a separation. This firmament between two areas of the

substrate: one is the world model, the stuff that we perceive as stuff in space perception, which Descartes later calls res extensa. And everything else is ideas, thoughts, reflections that are asynchronous to our perception, that happen independently of it, and they're

clearly separated, right? Because if you were perceiving your thoughts, you would be hallucinating them and you could not function. So you have a clear separation between the world and the ideas sphere. And in this all text, the world is called Earth, and the other one, Heaven, the

sphere of ideas. So the next thing is it, uh, figures out to create oscillations in there, and it takes the intensity of the oscillation as you see, it's brightness with the color of the day and the flatness of this, uh, contrast with dark. In this way, it now has a continuous

Dimension, and by combining dimensions, we can create objects. The first object it makes is by combining two dimensions: the plane. And the plane gets associated with the ground, and the baby happily crawls on the ground and makes sense of it in 2D. And at some point, it's

able to discover the third dimension, and now it can reason about how to build a tower, for instance, which very small babies cannot really understand. And then it figures out how to make solid objects in liquid and organic shapes and so on. And it understands how lighting changes

over time and becomes invariant against changes in lighting. And then it creates all the plants and the animals and gives them all their names. Right? Again, this is not physical objects, it's objects in space in your mind that have names. And then it's also creating a

model of the interest of the organism. And it realizes that the purpose of the exercise is to navigate the organism in the world. And then at some point, it identifies as the organism. And we see this in children typically happening between 2 and a half and 5 years old

that they talk about themselves in the first person. Before that, they can already talk and they talk about themselves in the third person. And I don't think that's because 'I' is such a complicated word or because they never hear anybody using it. It is, I think,

because they do not identify in the first person before they make this switch. Before the switch to this different spirit that is created in the image of the original Consciousness but is something that thinks of itself as man as woman, and woman, something that identifies as a human being. Right? So we

are basically created by our consciousness, by our infant baby consciousness in our mind, and then put into this world, in the simulation world that our mind is creating. And then we forget that we created this world in our own mind and identify as this human

being that is in this world and it's subjected to it. And at some point we can start meditating or we take acid and then wake up from this again and realize, oh my God, I'm actually not that self model. I'm not that self. That self is

just a model of a person that doesn't actually exist. It's a story that my mind is telling itself and I'm actually the universe. I'm everything. And then you wake up a little bit more and you realize, oh no, I'm actually the thing

that generates it all. I am the generator of everything that I'm interacting with. That makes sense to me, right? So what consciousness is basically creating is a very sophisticated natural learning algorithm, a self-perpetuating intelligent recurrent information transformer, abbreviated spirit. Spirit is basically organizing

software agent. Oh my God, we have just rediscovered something. There are software agents in nature. A software agent is a causal pattern that is able to change things but it's not physical, right? The software in this computer is not physical, but it's causal. It can change things and

it's stable. And if it's in nature, it needs to be self-stabilizing. It needs to entrain itself on the substrate and control the substrate so it can be implemented. And it's a very interesting perspective if we realize that the actual invariance in living things is

not the molecules of these cells or the mechanisms or the genes, but the software that is running on them. That's a very interesting perspective. It's basically animism.

To continue watching this video, click the link in the top left or in the description below, or visit IA.tv for more debates and talks from the world's leading thinkers on today's biggest ideas.

A radical theory of consciousness | AI researcher Joscha Bach

The Institute of Art and Ideas creates hundreds of educational debates, lectures, and interviews on Philosophy, Science, Politics, and Arts. 

Learn from world-leading thinkers on the big ideas. From Hitchens to Galloway, Penrose to Pigliucci, our experts will teach you feminism to foreign policy, and quantum physics to consciousness.

And if you want even more ideas from the world's greatest thinkers, visit our website at IAI.tv

Or listen to our podcast, Philosophy For Our Times. Links below!

A radical theory of consciousness | AI researcher Joscha Bach

A radical theory of consciousness | AI researcher Joscha Bach

Scripts