What is Creativity?
Nivi: I just want to chat about what is creativity. I’m not going to say what it is, because nobody knows what it is, but I will share some thoughts.
So there’s a lot that can happen in the universe without creativity, but there’s a lot more that can happen with creativity or conjecture.
If something is implicit in the Big Bang, you don’t need to invoke creativity to explain why that thing happened. Likewise, if something happens through the spontaneous action of the laws of physics, like an electron flying off of an atom, you also don’t need to invoke creativity to explain that.
The output of any deterministic computer program is also not creative, like all LLMs, even though they have pseudo-random inputs that go into them that make it seem like it’s coming up with something creative. If you control for the pseudo-random input, it’ll still give you the same answer every time. Also, it’s pseudo-random, not random, which just means even the randomness is a deterministic.
David Deutsch likes to invoke the terraforming of Manhattan as something that could never be explained through the implicit effects of the Big Bang or some spontaneous physical reaction. You have to invoke creativity to explain why Manhattan looks the way it is, or why there’s a statue of somebody on a corner somewhere.
The creativity of evolution
Now, it’s possible that you could get human level creativity through randomness. Evolution, for example, is creative and it’s driven through randomness. It’s not human level creativity, but it is creative. But the problem with evolution is that it’s accidentally making solutions randomly through variation and selection.
It’s a random process of mutations, and the DNA gets turned on and off through a host of means I won’t go into, and then transcribed into proteins which have functions like enzymes that catalyze reactions, and you end up with something like the wings of a bird or a human eye that was nowhere implicit in the Big Bang.
We do have an explanation for where the creativity in evolution comes from, and it’s the only other thing that we’re aware of that’s creative in the universe besides humans, which tells you how profound Darwin’s theory really is, that he was able to come up with an explanation for at least the creativity in evolution while we still don’t have an explanation for how humans come up with ideas or conjecture things.
But the problem with evolution is that it’s a random creative process that can’t be directed. So DNA or evolution doesn’t have problems that it’s trying to solve like a human being does. And when it does encounter a problem, it needs to have already accidentally solved it because the problem-solving process of mutation and evolution is random and extremely slow.
The knowledge also that it creates is not explanatory knowledge. It doesn’t explain anything. It does have the knowledge of how to build an eye encoded in it, but it doesn’t know how to answer questions and give explanations for why things happen.
It also can’t access the whole universe of knowledge because it can only reach points in the space of knowledge that go through paths consisting of viable organisms. Meaning the only way to get to a certain point in the space of knowledge is through offspring that survives. If a certain kind of creature with seven eyes would require a path through the space of knowledge that did not result in organisms that could live or even be created or survive for more than a split second, you wouldn’t be able to make a creature that has seven eyes . So there’s points in the space of knowledge that DNA cannot access.
It’s possible that there is some random or non-random low-level process of variation and selection that could create a higher level emergent process that is more willful with its outcomes. So it could have a problem, it can decide it wants to work on the problem, and then solve that problem. That’s purely speculative.
Creativity cannot be automated
Now I just want to share some philosophical thoughts on what is creativity more generally. So creativity is the part that hasn’t been automated. If you can explain creativity well enough to automate it, that process was actually never creative.
For example, we have LLMs that can make stories for us. Those stories are not creative because we have a deterministic explanation for how those stories are generated.
When I say deterministic, I mean a process where the outcome of the process is determined by an earlier state of the process. So the same inputs to the program will always give you the same output when you control for any pseudo-random numbers they’re putting in there.
So a person can write the same exact story as ChatGPT, and the human was creative in the process and ChatGPT could write the greatest story in the history of humanity, and it still would not have been a creative process. All of the creativity was in the prompt that it was provided and the creativity of the programmers.
And that’s because a calculator or an LLM can solve a problem for a human in a way that does not involve conjecture, whereas the person would’ve had to have used conjecture to solve the problem. We don’t look at the output of a process and say whether that was creative or not. The creativity is in the method or really in the lack of method. And the more you can explain a creative process and turn it into a deterministic computer program, the less that process that you are documenting was actually ever a creative process.
Creativity is not a deterministic mixing of it’s inputs
One thing that gets people confused is that creativity does have inputs, but it’s not a deterministic mixing of its inputs. So you do rely on the things you’ve heard or seen to come up with creative things, but it is not a deterministic mixing of its inputs where you can just put those inputs into a computer program and come up with the same output as a creative or conjectural process.
So creation makes knowledge, the Latin phrase is ex nihilo, however you pronounce it, E-X space N-I-H-I-L-O, which means out of nothing. The creative part or the conjectural part is the thing that was created out of nothing. We do rely on the past to create new knowledge, but if it was already determined by the past, there would be nothing new in it.
And I think the creative process is inexplicit, meaning that we don’t know how to put it into words yet. We know how to do it. We don’t know how to put it into words. A similar example of something that’s inexplicit is: We know how to ride bicycles, but we don’t really know how to put it into words, so that someone can read how to ride a bicycle and then do it.
The lack of deterministic automation behind the source of new knowledge is a good thing because understanding it that way seems like it would actually mean the end of the growth of knowledge, or not the complete end of the growth of knowledge, but it would put a bound on the growth of knowledge. The space of it would still be infinite, but it would still put a bound around it. Whereas if you can’t automate it, then you can always increase the scope of knowledge creation through new rules, for example, that can be made through a conjectural process.
So what we need is an explanation for creativity that doesn’t ruin the means of creativity, where the explanation cannot be turned into a set of steps where the outcome is determined by following the steps with a given input.
FInally, I want to note that anything that we consider to be creative, we consider to be alive. So whether that’s the creativity of evolution or the creativity of the human mind or the creativity of an AGI.
The multiverse provides a canvas for creativity
Okay. One last topic I briefly want to discuss how you can have creativity in a universe where our best explanations of physical phenomenon are all deterministic.
The technical term for that would be, they don’t have any counterfactual property, which means that, say the state of a bit in a transistor and the next state of the bit after that, and the state after that has already been determined by the evolution of the laws of physics. So there is no counterfactual property to the state of the bit and the transistor, which just means there’s no way it could have been different. What’s the point of even running the computer, other than to find the result, but you couldn’t use the computer to do anything other than to create the result.
Let me give you an example that is perhaps more relatable. If there were no counterfactual property to the universe, meaning things have to be the way they are, or to say it another way, they couldn’t be otherwise, every idea you have is going to be predetermined. It couldn’t have been that you had a different idea or a better idea, or that even that you could potentially improve that idea: that improvement is already predetermined.
So how do you get a counterfactual property in a deterministic universe so you can have the possibility that things happen and don’t happen which is required for creativity. If something is already determined to happen, I don’t consider that to be creative.
There needs to be something that has the free will to create it. So how do you get that? We don’t have a full explanation or even anything close to it, but I will just say one thing:
The way that you can get counterfactuals in a deterministic universe is through the quantum multiverse because there are always universes where the transistor is one, and there are universes where the transistor is zero. So there are universes where you had that great idea and there are universes where you didn’t have that great idea.
One of the requirements for something to be creative is that there has to be the possibility that it didn’t happen, so the quantum multiverse provides a canvas for creativity by providing universes where it didn’t happen.
Just to be clear, I’m not saying that the multiverse explains creativity. I’m just saying that it has the property that it allows for counterfactuals.
It’s interesting that the multiverse can do this in a way that doesn’t require randomness and is deterministic because there is a deterministic function they call the wave function that operates on the multiverse and describes the evolution of the multiverse as a whole, deterministically. From the point of view of an individual in a specific history things can still look random from your point of view, but the whole of the universe is evolving in a deterministic way.
I also want to note that it’s interesting that the histories of the multiverse differentiate not just from spontaneous processes, like an electron popping off of an atom somewhere, but they also differentiate through creative processes like someone’s idea spreading across some histories of the multiverse and not others.
That’s another way that humans are a force of nature because just like an electron either popping off or not popping off of an atom will create differentiation of universes, you having an idea can create differentiation of the multiverse, which to me seems very cool.
Q&A
So anybody have any questions? Maybe just raise your hand and I’ll give you the mic.
Question: Greetings from El Salvador. I am a playful maxi. I’m a maxi regarding play. I’ve studied play science for quite a few years and worked with the National Institute for Play and have IP in that particular arena as well. My question to you is, what is the role of play and it’s, as far as I know, something that can’t be easily quantified or measured, certainly not deterministic, and I would advocate that is the key to that component that seems to be the mystery right here.
Nivi: Good question. I don’t have the answer. I think you probably know more about it than I do. I think it’s a good idea. David Deutsch talks about the fun criterion, which is finding things that you think are fun and working on that and using that as a criterion for what to work on. I’m probably misquoting him and it’s probably got finer details than that, so maybe look into that if you like.
Which you probably already know about, but just in case you don’t or maybe other people don’t, and then we also know about the school of taking children seriously, which I haven’t studied but is worth investigating. And Brian Norgard has a good tweet about play where he says, “Your curiosity is the compass”.
Anybody have any questions?
Question: You said that being able to automate creativity will put a bound on how much knowledge we can create or something along those lines. I got confused about that because my understanding was being able to automate creativity would mean being able to figure out how we’re able to conjecture and make new guesses. How does that put a bound on what knowledge we can create.
Nivi: There’s a distinction between explaining something and automating it. So we want an explanation for creativity. An automation of creativity will never be creative. So an automation of creativity would be a computer program or a clear set of steps that you can follow to transform an input to an output, that would be automation.
That’s different than explanation. We have an explanation for evolution, through variation and selection that Darwin came up with. But we don’t have a way to automate that, and I’m not even sure what it would actually mean to automate that. We do have evolutionary algorithms, but I’m not sure they really amount to anything. And they certainly have never created life, and it’s not even clear that they’ve ever actually created new knowledge. You can look into David Deutsch’s critiques of that if you want, of evolutionary algorithms. So I would distinguish between automation and explanation.
But there was also another part of your question that I forget. Can you repeat it for me?
Question: What’s the reasoning behind saying automating creativity will put a bound on the amount of knowledge we’re able to create?
Nivi: I probably explain it better in the first episode of the podcast, but essentially if you create a language to express something then all of the expressions that that language can express set a bound on what can be created. That’s all you can make. There’s a finite set of things that you can say in English. Not finite, it’s infinite, but it’s still bounded. You can have a infinite set that’s bounded. You can have a bigger infinite set that you can’t get to.
For example, in English, the way we get around that is you can go and redefine all the words with new ideas and use that to come up with new ideas. Or you can put the words in different orders that break the rules and use that to come up with new ideas, or create new words, or leave the idea of language altogether and start incorporating dance and music into it.
That’s how you can take a formal system and increase its scope. Whereas if you put creativity into a formal system, its scope would be bounded, it would be still be an infinite set, but it’d be bounded.
Question: Would it be something like the Cantgotu environments that David talks about in his book?
Nivi: That is also another way that you can increase the size of a formal system. I gave you some simple ways like change the definitions of the words. But you could find a statement that is neither true nor false in the formal system, and then add that statement to the formal system as a truth or a falsehood. Take your pick, doesn’t matter. And that’s another way to extend a formal system. I think mathematicians do that, when they take an undecidable proposition and add it as an axiom.
I appreciate you guys taking the time and thought to come up with a question because it helps me think about things more.
Cheers. I’ll try to do another one soon. Thanks everybody for hanging out.