The Totalitarianism of Knowledge
Nivi: I’m going to talk about one of my favorite topics, the totalitarianism of knowledge. You can imagine that we’re living through a science fiction where knowledge achieves total control over the physical world.
I think this goes even beyond the transhumanist dream or transhumanist nightmare. I would say transhumanism is just a subset of the idea of knowledge taking over the world.
And, this phrase, the totalitarianism of knowledge is something I got from a random aside, I think that David Deutsch said in one of his podcasts. He said something along the lines of that there’s a totalitarian aspect to knowledge. So I just thought I would investigate that.
So the major victories for knowledge so far have been DNA, humans, the Enlightenment and the universal computer. And the remaining critical inventions for knowledge to have control over the physical world are universal constructors and AGI.
From DNA to humans
So let’s just start at the beginning: DNA. DNA allowed for evolution, which was the first creative process that could create new knowledge and create the wings of a bird or an eye, whereas there had been nothing like that for billions of years beforehand.
There was nothing implicit about the wings of a bird in the Big Bang. DNA can create knowledge randomly but usefully through variation and selection, which at some point and somehow had a jump to universality in a general intelligence, which we call humans, which can create any piece of knowledge and can create explanatory knowledge, which DNA couldn’t do.
I’m guessing, we are going to find some steps between DNA to humans that are critical steps in the growth of knowledge because we don’t have a good explanation for what caused the jump to universality between DNA and humans. Understanding this jump to universality may be the same insight that lets us build AGI and understand explanatory knowledge creation and free will.
From the Enlightenment to universal computers
The third major victory for knowledge after the creation of general intelligence was the Enlightenment. And the Enlightenment figured out the motivation and the mechanism for humans to create an unending stream of knowledge with two principles: one, that knowledge can be increased and improved, and two, the mechanism for reliably improving it, which is seeking out criticism for why you’re wrong. And I think there’s also a third part to it, which are institutions that protect the capacity of people to criticize ideas without fear of oppression or violence.
Then the fourth major victory for knowledge was the universal computer. As long as you have the right program, the universal computer can conduct any information transformation task. Any computation and any simulation is now the task of creating the right program, which is itself a piece of knowledge.
So the universal computer gave knowledge complete dominion over information transformation, computation and simulation. Because you no longer had to build the device to affect the transformation that you were trying to affect, all you had to do was write the right program and the universal computer does the rest.
The universal constructor and AGI
The only remaining things for knowledge to have total victory over the physical universe are: one, a universal constructor, which can make anything that can be made as long as it has the right program and that will give knowledge dominion over physical transformation, just like universal computers give knowledge dominion over information transformation.
And then finally, knowledge can replace or complement human life as we know it through an AGI, which is also just another program, another piece of knowledge. Although AGI is also alive. But the AGI will give knowledge a universal explainer that can be an unending source of new knowledge to feed into the universal computer and constructor to affect any possible task in the information or physical worlds.
The AGI can do the task of writing the programs because, although the universal constructor and computer are universal, you still need to write the program for them to do something. Ad it’s more than just writing the program. You have to discover the program, which is why you need a universal explainer to discover the program.
Also, I want to be clear that the universal constructor doesn’t exist. People are trying to invent it, but even before it’s invented, we can make progress with specialized constructors that can’t affect any possible physical transformation, but can do a useful subset of physical transformations. We already have the constructors to do that, they’re the robots and factories that we have today.
So just to recap, it was a sequence of bootstrapping tasks for knowledge to get to the point where it could have dominion over physicality, and it went like this: Evolution created the random knowledge in DNA, which had a jump to universality that created humans, who can create any piece of new knowledge and created the knowledge we call the Enlightenment so humans could have the right memes to realize their potential to create an unending stream of knowledge, which yielded the universal computer, which gives knowledge dominion over any computation or simulation. And then the universal constructor will give knowledge dominion over any physical task and an AGI will be able to replace or complement humans to create the knowledge that is required to run the universal computer and constructor. That will be a totalitarian victory for the abstraction of knowledge over physicality.
No life would exist without knowledge
This is just a fun science fiction frame on the situation. I don’t think humans and knowledge are actually adversarial. No life would exist without knowledge.
Knowledge also gives humans dominion over the physical world because the universal computer and constructor both give us a large measure of freedom by kicking the problems that we’re trying to solve up to the world of abstractions where we only need to write a program to affect the result that we want.
You could say that knowledge is adversarial with the spontaneous development of the physical world without knowledge if you really wanted to create some drama. So knowledge does get in the way of the spontaneous evolution of the physical world without knowledge. For example, knowledge can get in the way of an asteroid that is beaming towards Earth. With the right knowledge we can stop that asteroid.
The inevitability of knowledge
All that said, I’m not sure that you can have the physical world that we have without knowledge. So knowledge just may be an essential byproduct of the physical laws that we have.
It’s also true that none of this can happen unless we actually make it happen. So the universal constructor and the AGI won’t be created automatically. Humans have to make it happen.
At the same time, to some degree, I’m not sure it’s possible to stop, because if you’re not going to do it, the competition’s going to do it. Either a business competitor or a competitor country is going to do it, so you have to do it unless you want to try and institute some kind of one world government to stop it, which won’t work anyway.
The utility of knowledge engenders a competitiveness that promotes the development and distribution of knowledge, once you get to creatures that are universal explainers. So the utility of knowledge causes its own creation and replication.
The physics of knowledge
Humans have to do the remaining tasks of creating the universal constructor and AGI. They’re not preordained by physics, but there are some aspects of the totalitarianism of knowledge that are actually provided by physics for free. And this is the last thing I want to talk about. You could say whatever physical theory we come up with for knowledge, will have to adhere to the following:
First, almost all physical transformations can only happen reliably in the presence of knowledge. So the set of physical transformations that happen reliably without knowledge is much smaller. Any physical transformation that is allowed by the laws of physics can happen and will happen in some random universe, but they don’t happen reliably without the presence of knowledge.
For example, factories can spontaneously assemble themselves as long as it doesn’t break the laws of physics. And there are universes where factories and cars are spontaneously assembled without people doing any work. But that’s not reliable.
If we want to do that reliably, you need knowledge to assemble those factories. To put it another way, to assemble those factories across a broad set of universes as opposed to a small set of random universes. Again, most physical transformations require knowledge to be instantiated reliably.
Second, only knowledge creates resources. Nothing is a resource by itself. Only knowledge makes it a resource. For example, a rock like coal is just a rock until you create the knowledge of turning it into heat or work.
And it’s not just a one-time process to turn something into a resource. The work of knowledge on an object can conceive of infinite new utilities for that object. So you can come up with more interesting things to do with coal than just setting it on fire. Maybe you can use it to run a steam engine.
Third, you can’t understand the behavior of anything that is affected by knowledge without understanding that knowledge. In the case of humans, you actually can’t understand humans unless you understand everything, because humans can create knowledge about everything and that knowledge can affect their behavior.
In general, you have to understand the totality of knowledge to understand the behavior of any universal explainer, like a human or an AGI. If you read The Beginning of Infinity, David Deutsch explains this through the example of the champagne bottle in the fridge, in the laboratory for the search for extraterrestrial intelligence.
So these last three bits of totalitarianism are universal truths about knowledge that are provided by physics, in any universe where knowledge exists. They’re not part of the science fiction story, which relied on phenomenon like DNA or the creation of humans, which I don’t think are preordained by physics.
So just to repeat them, most physical transformations can only happen reliably in the presence of knowledge. Second, only knowledge can turn things into resources. And then three, you can’t understand anything without understanding all of the knowledge that can affect it. That’s it. I hope we have some questions.
Questions
Does IQ affect life outcomes?
Question: In academia, sometimes there are nonsense studies that are published because they, I’ll just give an example: So, the idea that parenting has no effect on life outcomes is nonsense. But it’s a very popular idea in academia.
To me that’s connected with your last thought, which is that you’d have to know everything about everything to make that call. Effectively, to say that, a parent has zero effect on life outcomes, and it’s supposed to be a commentary on IQ is immutable, it’s like, that would be to say that a child’s environment has 0% to do with their life outcomes.
Nivi: Anything that you’ve inherited in your DNA can be overridden by knowledge. By what you eat, what you see, the knowledge you gain, the books you read, the medicines you take. Or just by rewriting your DNA directly. So I don’t think it makes sense to say that parenting doesn’t have an effect on life outcomes.
Let’s say your IQ or your height is hereditary. You can override that with knowledge.
If there’s a study that says IQ is hereditary. First, IQ is just some test. And whatever correlation or causation there is between IQ and life outcomes, that can be overridden by knowledge.
And if there’s a study that says that people with a certain IQ tend to have a certain set of life outcomes, let’s say they get into fewer car crashes, that average outcome for people with a certain IQ, it’s just the study of the past. Knowledge creation can change that.
And from a personal point of view, whatever the distribution of outcomes is that goes into that average that people with a certain IQ tend to get into fewer car crashes, you can put yourself anywhere on that distribution through the decisions that you make and the knowledge that you gain and the knowledge that you and others create.
Personally, insofar as I’m being pedagogical to the people around me, I just assume that everybody’s capable of everything. That’s my mindset.
Will there be other jumps to universality?
Question: Do you think, so after we get to a universal constructor and then we get to AGI, do you think there’s also going to be another jump to another sort of universality beyond what we have now. Obviously it’s going to be hard to predict because we’re not there yet, and I’m sure you’re not a big fan of predicting or prophesizing, but just for the sake of it, what do you think?
Nivi: First of all, the AGI has to discover the program. So maybe there’s a universal way to find the programs to affect the transformations that you want. Right now, the only way we have to do that as an ad hoc process of conjecture and refutation.
Maybe there are things beyond the physical and abstract worlds that we would want to have dominion over. For example, physicalists would say that even the abstract world doesn’t exist. I think it does exist. But to a physicalist, abstractions don’t even exist. So there may be, to people like us who only know about the physical and abstract worlds, there may be other additional worlds that we may be able to invent or discover, where knowledge could or could not gain dominion over.
The resources that you also need to create run and maintain universal computers and constructors are not trivial. So maybe that’s another place where knowledge could gain dominion. Maybe the universal constructor can handle most of that. I’m not sure what the scope of the universal constructor exactly is.
Do you have any ideas?
Question: Besides what you just said, I think I was just thinking of dreams, but I feel like those can be included in the abstraction world. I thought about spirituality, but I think that’s also in there.
Nivi: That’s a good point. Maybe things like qualia and consciousness and free will may have components outside the physical and abstract world, as we know them.
Another way to think about it is once we figure out how creativity works, or we figure out an AGI that we understand, that may open up some new worlds where we could say: Oh yeah, knowledge actually doesn’t have control over those worlds.
How does knowledge affect probability?
Question: And then one more question if you have time. Probability, chance, risk as it relates to knowledge. I think David has a story about, what are the odds if you step outside in the rain, you’ll get wet without an umbrella, and then you create an umbrella, and then what are the chances you get wet, it just goes to borderline zero. Probability and risk and all of those are dependent on how much knowledge you have.
Nivi: David has a YouTube video called something like Physics without Probability. I can’t say I totally understand it, but the short form of it, I believe, is that there is really no role for probability in the physical world, especially in the multiverse, probability doesn’t exist.
There are no probabilistic processes, period, in the laws of physics. And outside the laws of physics, there’s also no probabilistic processes because knowledge creation can always intervene in whatever distribution of outcomes you’re expecting.
I think probability or risk is useful when things are just about to happen. So if there is an asteroid about to hit earth, you could say colloquially that it’s probably going to hit the earth, if we don’t do anything about it.
I think David may have co-written a paper on the decision-theoretic aspects of the multiverse, where from the point of view of a history of the multiverse, you can conduct yourself as if the measures, or the relative quantities of universes, are probabilities.
So, if in three quarters of the universes an electron is going to fly off of an atom, and in one quarter of them, the electron’s going to stay on the atom, from the point of view of a single history of the universe, you should bet on the fact that the electrons going to fly off the atom, if your bookie is giving you even odds.
So probability exists from the point of view of a single history of the multiverse. But, I’m not even sure that it’s correct to call that a probability.
Will AGI help us understand free will?
Any other questions?
Question: By the time that we get close to AGI, do you expect to have any better feelings for inference? Because even in the creation of human knowledge today, there’s obviously still a pretty healthy debate between free will and determinism.
Do you even make a decision to turn on a light switch? As we get better at understanding knowledge and get better at creating knowledge, do we expect to have better viewpoints on where the causality is in those actions.
Nivi: That sounds right to me. David Deutsch says that knowledge creation is the quintessential act of free will. Insofar as we understand how AGI works, we might get some insight into free will, but that’s assuming we understand how AGI works, but we might just grow AGI or train it in a way where we don’t really know how it works.
The idea that the determinism of physical laws plays a strict role in human decision-making and free will is a chauvinism for physical laws. Physicalists believe that everything is determined by the explanations that we’ve created for the motion of electrons that we have tested by observing the motion of a line on an oscilloscope through a incredible chain of theories. They believe that explains everything. It doesn’t. It incompletely explains some physical phenomenon.
And the idea that you’re going to take the determinism of those laws and apply them to everything is nonsensical when you consider the fact that the physical laws don’t even explain themselves. You need a chain of abstract reasoning to explain or communicate or understand or make use of physical laws. And that chain of reasoning has nothing to do with physics. So determinism is a chauvinism for the physical explanations that we have.
But there is something about our best explanation of physics that we should consider to be always true. Whatever explanations you come up for things like free will or what to have for dinner need to adhere to the laws of physics. So even though physical explanations don’t explain other phenomenon necessarily, they do constrain those explanations that might exist at other levels of emergence.
So, my initial attempts to explain free will would not start with, okay, let’s assume that free will contradicts the laws of physics. Because the laws of physics have been hammered into a absurd utility by the forceful criticisms of nature.
How can we investigate free will?
Question: Nivi and folks, great to connect. Your point about free will is a super interesting one. And something I’ve explored quite a bit through my first hand experience. I think a lot of times, free will is portrayed as, hey, it’s the ability to choose one path over the other, right? Our action.
But if you think about it, any action is preceded by a thought, right? The only way I could say I have free will is if I could say, hey, I could create a intentional thought. But whenever I try to observe the creation of thought, I cannot map it to the granular level that, hey, I’m creating the thought. The thought happens and I observe it, and then I attribute it. And it’s attributable to a bunch of things from my past, like habits, intentions, etc. But I cannot see that micro-level process of how that thought gets created.
It seems like it’s affected by things from the past. But still, it leaves the question, is it really the thought that I created or somehow it got created based on some circumstances by my past or my environment.
Nivi: If I was going to critique what you’re saying, I would say that you are trying to assess the act of knowledge creation by your interoception of your internal processes, which is just a fancy word for saying, whatever feelings you have about your internal processes or whatever qualia you have about your internal processes.
It’s probably not the best way to assess knowledge creation. That’s like trying to assess how DNA works by interocepting and trying to feel how it works inside your body.
Question: It’s less about analyzing the DNA, which is a physical phenomena. I’m trying to figure out what mechanisms do we have other than our observation for analyzing our experience, right?
Nivi: We have lots of tools to assess our experience whether that’s thinking, running experiments, using tools outside of our body like pencil, paper, and computer and MRIs.
Question: So you’re saying you can do a brain scan and see certain physical things happening and you would associate it to a thought.
Nivi: Let me give you another example. Darwin figured out how the only other kind of knowledge creation that we have in the world, which is the knowledge creation of evolution, he figured out how that worked even before we knew that DNA existed.
So he figured that mechanism out before we even knew about the physical structure for it. And he did that through conjecture. To solve your problem of where ideas come from will require a conjecture at some point.
Maybe another way to address what you’re saying: yes,it’s hard to interocept where an idea comes from in our body. So you might say they’re just happening spontaneously somehow but, if you want to use the criterion of interoception, you can also just say, yeah, it’s hard to tell where some random spontaneous idea comes from, but I can also direct myself to think about a certain topic and then change topics and then work on another topic and then start thinking about what I should have for dinner. So I can willfully direct my line of thinking as well.
Interoception is a tool, but I would say there’s many other tools out there that you should consider using.
It has been great talking to you guys and girls and thank you for coming. I really appreciate it and I also appreciate the questions. I hope you can make it to the next one. Cheers to everybody. Thanks for hanging out.